From john.leidel at gmail.com Wed Nov 1 04:21:17 2006 From: john.leidel at gmail.com (John Leidel) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] SC'06 Beowulf Bash ? In-Reply-To: <200611011815.22799.csamuel@vpac.org> References: <200611011815.22799.csamuel@vpac.org> Message-ID: <27f776af0611010421k352068a6tddf69968f8f4c1d6@mail.gmail.com> What ever happened to the annual Cigar and Cognac party as well??? *Douglas Eadline* On 11/1/06, Chris Samuel wrote: > > Hi folks, > > I'm off to my first SC'06 this year to help man the VPAC booth with a > couple > of friends and was wondering if there were any details about the Beowulf > bash > yet ? > > cheers, > Chris > -- > Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager > Victorian Partnership for Advanced Computing http://www.vpac.org/ > Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia > > > > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.scyld.com/pipermail/beowulf/attachments/20061101/29b869bd/attachment.html From lusk at mcs.anl.gov Wed Nov 1 07:05:50 2006 From: lusk at mcs.anl.gov (Rusty Lusk) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] ch_p4mpd device problem In-Reply-To: <1162287893.19656.53.camel@sysadm.stp.cdac.ernet.in> References: <1162287893.19656.53.camel@sysadm.stp.cdac.ernet.in> Message-ID: <3B6421EB-0E4E-4BD9-9B79-FA70107D4D7A@mcs.anl.gov> I would recommend that you switch to MPICH2, which is the current version of our MPICH work. Regards, Rusty Lusk On Oct 31, 2006, at 3:44 AMOctober, ajit mote wrote: > HI , > i installed mpich-1.2.6 with ch_p4mpd device and started the > deamons ... > upto this point no problem , rung form very well... > but when i run the job it gives me error : > p1_13343: (2.019287) p4_peer_msg_handler: failed > connect > to 127.0.0.1 > p1_13343: p4_error: failed to connnect: -1 > p4_error: latest msg from perror: Connection refused > > where is the problem? ... > > > thanks ... > > -- > "Live Life Dangerously" > > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > From deadline at clustermonkey.net Wed Nov 1 10:15:14 2006 From: deadline at clustermonkey.net (Douglas Eadline) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] SC'06 Beowulf Bash ? In-Reply-To: <200611011815.22799.csamuel@vpac.org> References: <200611011815.22799.csamuel@vpac.org> Message-ID: <60043.192.168.1.1.1162404914.squirrel@mail.eadline.org> Beobash details: http://www.clustermonkey.net//content/view/166/2/ -- Doug > Hi folks, > > I'm off to my first SC'06 this year to help man the VPAC booth with a > couple > of friends and was wondering if there were any details about the Beowulf > bash > yet ? > > cheers, > Chris > -- > Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager > Victorian Partnership for Advanced Computing http://www.vpac.org/ > Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia > > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > -- Doug From deadline at eadline.org Wed Nov 1 10:21:44 2006 From: deadline at eadline.org (Douglas Eadline) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] SC'06 Beowulf Bash ? In-Reply-To: <27f776af0611010421k352068a6tddf69968f8f4c1d6@mail.gmail.com> References: <200611011815.22799.csamuel@vpac.org> <27f776af0611010421k352068a6tddf69968f8f4c1d6@mail.gmail.com> Message-ID: <46436.192.168.1.1.1162405304.squirrel@mail.eadline.org> In true LECCIBG fashion we are planning ahead this year. Which means we should have everything figured out by the end of next week. I'll announce the exact time and date as soon as we know it. I am thinking that Monday night after the Gayla should be a good time. Keep an eye on the LECCIBG page as well: http://www.clustermonkey.net//content/view/164/57/ -- Doug > What ever happened to the annual Cigar and Cognac party as well??? > *Douglas > Eadline* > > On 11/1/06, Chris Samuel wrote: >> >> Hi folks, >> >> I'm off to my first SC'06 this year to help man the VPAC booth with a >> couple >> of friends and was wondering if there were any details about the Beowulf >> bash >> yet ? >> >> cheers, >> Chris >> -- >> Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager >> Victorian Partnership for Advanced Computing http://www.vpac.org/ >> Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia >> >> >> >> _______________________________________________ >> Beowulf mailing list, Beowulf@beowulf.org >> To change your subscription (digest mode or unsubscribe) visit >> http://www.beowulf.org/mailman/listinfo/beowulf >> >> >> >> > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > -- Doug From deadline at clustermonkey.net Wed Nov 1 10:23:19 2006 From: deadline at clustermonkey.net (Douglas Eadline) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] SC'06 Beowulf Bash ? LECCIBG In-Reply-To: <27f776af0611010421k352068a6tddf69968f8f4c1d6@mail.gmail.com> References: <200611011815.22799.csamuel@vpac.org> <27f776af0611010421k352068a6tddf69968f8f4c1d6@mail.gmail.com> Message-ID: <39845.192.168.1.1.1162405399.squirrel@mail.eadline.org> In true LECCIBG fashion we are planning ahead this year. Which means we should have everything figured out by the end of next week. I'll announce the exact time and date as soon as we know it. I am thinking that Monday night after the Gayla should be a good time. Keep an eye on the LECCIBG page as well: http://www.clustermonkey.net//content/view/164/57/ -- Doug > What ever happened to the annual Cigar and Cognac party as well??? > *Douglas > Eadline* > > On 11/1/06, Chris Samuel wrote: >> >> Hi folks, >> >> I'm off to my first SC'06 this year to help man the VPAC booth with a >> couple >> of friends and was wondering if there were any details about the Beowulf >> bash >> yet ? >> >> cheers, >> Chris >> -- >> Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager >> Victorian Partnership for Advanced Computing http://www.vpac.org/ >> Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia >> >> >> >> _______________________________________________ >> Beowulf mailing list, Beowulf@beowulf.org >> To change your subscription (digest mode or unsubscribe) visit >> http://www.beowulf.org/mailman/listinfo/beowulf >> >> >> >> > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > -- Doug From becker at scyld.com Wed Nov 1 15:19:54 2006 From: becker at scyld.com (Donald Becker) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] SC'06 Beowulf Bash ? In-Reply-To: <200611011815.22799.csamuel@vpac.org> Message-ID: On Wed, 1 Nov 2006, Chris Samuel wrote: > I'm off to my first SC'06 this year to help man the VPAC booth with a couple > of friends and was wondering if there were any details about the Beowulf bash > yet ? I'm the one holding up the formal announcement. There was a plan to hold a joint meeting of the BayBUG and BWBUG (west coast and east coast Beowulf User Groups) before the bash, and I was suppost to find people for a panel. I'm now trying to convince people that it's better for the meeting to be the bash itself, and the topic to be "Beer: is more always better?" I'm pretty certain that we can find speakers for that. BTW, We are still accepting sponsors. (More than "accepting", how 'bout "would be really happy to have a few more".) Contact "suzie" at beowulf dot org if you or someone you know would be interested. -- Donald Becker becker@scyld.com Scyld Software Scyld Beowulf cluster systems 914 Bay Ridge Road, Suite 220 www.scyld.com Annapolis MD 21403 410-990-9993 From csamuel at vpac.org Wed Nov 1 19:35:39 2006 From: csamuel at vpac.org (Chris Samuel) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] SC'06 Beowulf Bash ? In-Reply-To: References: Message-ID: <200611021435.40074.csamuel@vpac.org> On Thursday 02 November 2006 10:19, Donald Becker wrote: > I'm now trying to convince people that it's better for the meeting to > be the bash itself, and the topic to be "Beer: is more always better?" > I'm pretty certain that we can find speakers for that. It depends on the beer. 'nuff said.. :-) Brains S.A., er, Chris.. -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 191 bytes Desc: not available Url : http://www.scyld.com/pipermail/beowulf/attachments/20061102/46b6e51a/attachment.bin From nixon at nsc.liu.se Thu Nov 2 00:03:41 2006 From: nixon at nsc.liu.se (Leif Nixon) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] SC'06 Beowulf Bash ? In-Reply-To: <200611021435.40074.csamuel@vpac.org> (Chris Samuel's message of "Thu, 2 Nov 2006 14:35:39 +1100") References: <200611021435.40074.csamuel@vpac.org> Message-ID: Chris Samuel writes: > It depends on the beer. 'nuff said.. :-) > > Brains S.A., er, Chris.. I had a real problem with that at CCGrid in Cardiff. How are you supposed to order a pint of Brains with a straight face? -- Leif Nixon - Systems expert ------------------------------------------------------------ National Supercomputer Centre - Linkoping University ------------------------------------------------------------ From csamuel at vpac.org Thu Nov 2 03:37:28 2006 From: csamuel at vpac.org (Chris Samuel) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] SC'06 Beowulf Bash ? In-Reply-To: References: <200611021435.40074.csamuel@vpac.org> Message-ID: <200611022237.28719.csamuel@vpac.org> On Thursday 02 November 2006 7:03 pm, Leif Nixon wrote: > Chris Samuel writes: > > It depends on the beer. ?'nuff said.. :-) > > > > Brains S.A., er, Chris.. > > I had a real problem with that at CCGrid in Cardiff. How are > you supposed to order a pint of Brains with a straight face? I guess being a native of Cardiff helps (you learn from osmosis around pint #3 that it was founded by Samuel Arthur Brain [1]). Still, there's good reason that some folks believe the S.A. stands for Skull Attack[2]. ;-) I've been desperately thinking of a way to get this back onto clusters with no luck so I'll give up now before anyone notices.. Pob hwyl! Chris [1] - http://www.sabrain.com/index.cfm?alias=company [2] - http://en.wikipedia.org/wiki/S_A_Brain -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia From landman at scalableinformatics.com Sun Nov 5 05:54:33 2006 From: landman at scalableinformatics.com (Joe Landman) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Thought that this might be of interest Message-ID: <454DED19.3060300@scalableinformatics.com> We are working on a machine with 2 Opteron 2218s. For laughs, I ran streams on it. Here are 1,2,3,4 processor data. 1 processor: Copy: 5713.9944 0.0591 0.0560 0.0600 Scale: 5713.9822 0.0587 0.0560 0.0600 Add: 5454.2389 0.0911 0.0880 0.0920 Triad: 5454.1576 0.0916 0.0880 0.0920 2 processors: Function Rate (MB/s) Avg time Min time Max time Copy: 5713.8788 0.0591 0.0560 0.0640 Scale: 5713.9822 0.0596 0.0560 0.0640 Add: 5713.9538 0.0858 0.0840 0.0880 Triad: 5713.9538 0.0871 0.0840 0.0880 3 processors: Copy: 8888.4000 0.0391 0.0360 0.0400 Scale: 8888.4147 0.0391 0.0360 0.0400 Add: 8570.9733 0.0573 0.0560 0.0600 Triad: 8570.9551 0.0587 0.0560 0.0600 4 processors: Copy: 11428.1834 0.0293 0.0280 0.0320 Scale: 11428.1347 0.0298 0.0280 0.0320 Add: 10908.5812 0.0449 0.0440 0.0480 Triad: 10908.6107 0.0440 0.0440 0.0440 I did get one outlier at 4 CPUs. Could be a timing glitch. 4 processors: Copy: 13332.7765 0.0302 0.0240 0.0360 Scale: 13332.7765 0.0262 0.0240 0.0320 Add: 13332.6220 0.0400 0.0360 0.0480 Triad: 14999.0756 0.0391 0.0320 0.0480 This is a dual socket 1207 motherboard. DDR2/533 ECC Registered memory (will look back and check to be sure). Initial benchmarks (2.6 GHz clock) put it at about 17% faster than Opteron 275 and Woodcrest 5150 (2.66 GHz) on a GAMESS test we use (1h41m for Woodcrest and 275, and 1h26m for this unit). Joe -- Joseph Landman, Ph.D Founder and CEO Scalable Informatics LLC, email: landman@scalableinformatics.com web : http://www.scalableinformatics.com phone: +1 734 786 8423 fax : +1 734 786 8452 cell : +1 734 612 4615 From diep at xs4all.nl Sun Nov 5 14:37:52 2006 From: diep at xs4all.nl (Vincent Diepeveen) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Thought that this might be of interest References: <454DED19.3060300@scalableinformatics.com> Message-ID: <005701c7012b$0a460150$9600000a@gourmandises> Impressive bandwidth. Is this the so called 'k8h' revision which has improved SSE2 throughput? Mind running a prime95 test at 1 core and then at 4 cores, and print the iteration time you can get, like you can find it here? http://www.mersenne.org/bench.htm That's the time to do a multiplication and a modulo if i understand well in order to try to win $100k : http://www.eff.org/awards/coop.php Thanks, Vincent ----- Original Message ----- From: "Joe Landman" To: "beowulf" Sent: Sunday, November 05, 2006 2:54 PM Subject: [Beowulf] Thought that this might be of interest > We are working on a machine with 2 Opteron 2218s. For laughs, I ran > streams on it. Here are 1,2,3,4 processor data. > > 1 processor: > > Copy: 5713.9944 0.0591 0.0560 0.0600 > Scale: 5713.9822 0.0587 0.0560 0.0600 > Add: 5454.2389 0.0911 0.0880 0.0920 > Triad: 5454.1576 0.0916 0.0880 0.0920 > > > 2 processors: > > Function Rate (MB/s) Avg time Min time Max time > Copy: 5713.8788 0.0591 0.0560 0.0640 > Scale: 5713.9822 0.0596 0.0560 0.0640 > Add: 5713.9538 0.0858 0.0840 0.0880 > Triad: 5713.9538 0.0871 0.0840 0.0880 > > > 3 processors: > > Copy: 8888.4000 0.0391 0.0360 0.0400 > Scale: 8888.4147 0.0391 0.0360 0.0400 > Add: 8570.9733 0.0573 0.0560 0.0600 > Triad: 8570.9551 0.0587 0.0560 0.0600 > > > 4 processors: > > Copy: 11428.1834 0.0293 0.0280 0.0320 > Scale: 11428.1347 0.0298 0.0280 0.0320 > Add: 10908.5812 0.0449 0.0440 0.0480 > Triad: 10908.6107 0.0440 0.0440 0.0440 > > > > I did get one outlier at 4 CPUs. Could be a timing glitch. > > 4 processors: > Copy: 13332.7765 0.0302 0.0240 0.0360 > Scale: 13332.7765 0.0262 0.0240 0.0320 > Add: 13332.6220 0.0400 0.0360 0.0480 > Triad: 14999.0756 0.0391 0.0320 0.0480 > > This is a dual socket 1207 motherboard. DDR2/533 ECC Registered memory > (will look back and check to be sure). > > Initial benchmarks (2.6 GHz clock) put it at about 17% faster than Opteron > 275 and Woodcrest 5150 (2.66 GHz) on a GAMESS test we use (1h41m for > Woodcrest and 275, and 1h26m for this unit). > > Joe > > -- > Joseph Landman, Ph.D > Founder and CEO > Scalable Informatics LLC, > email: landman@scalableinformatics.com > web : http://www.scalableinformatics.com > phone: +1 734 786 8423 > fax : +1 734 786 8452 > cell : +1 734 612 4615 > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > From landman at scalableinformatics.com Sun Nov 5 15:38:25 2006 From: landman at scalableinformatics.com (Joe Landman) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <005701c7012b$0a460150$9600000a@gourmandises> References: <454DED19.3060300@scalableinformatics.com> <005701c7012b$0a460150$9600000a@gourmandises> Message-ID: <454E75F1.4010202@scalableinformatics.com> Vincent Diepeveen wrote: > Impressive bandwidth. > > Is this the so called 'k8h' revision which has improved SSE2 throughput? As far as I know the k8h is a quad core (found by googling). SSE throughput does appear to be better simply running a code that performed less well on a 275, but the performance difference (17.6%) was IMO indistinguishable from a clock speed delta (18.2%) to within the measurement error ... that is, if SSE2 does in fact run faster, I have not probed it yet. Using the PGI and Intel compilers. The latter appears to be doing its old tricks: I get > landman@jackrabbit:~/URS$ ./xxxxxx.woodcrest > > Fatal Error : This program was not built to run on the processor in your system. Uh huh. I took that string and googled it. http://www.devx.com/amd/Article/28001 If I had time I would no-op that test out and see how well it does. The effort to do this is larger than I care to do. Well, if anyone has a quick set of strings to look for, I really don't want to spend the time tracing the startup of the code.... I can insert a few 0x90 by hand. Its a shame as the late model Intel compilers generate reasonably good code. Since they wish to do it only for Intel processors, and the world is decidedly mixed, this has implications on the use of Intel compilers for lots of people wishing to get the best performance on all platforms with a single compiler tool. Doesn't it. I thought they had fixed that in the 9.0 compilers. Looks like I was wrong. > > Mind running a prime95 test at 1 core and then at 4 cores, and print the > iteration time you can get, > like you can find it here? > > http://www.mersenne.org/bench.htm Will pull it down, and see if I can get it going in under 10 minutes. > > That's the time to do a multiplication and a modulo if i understand well > in order to try to win > $100k : http://www.eff.org/awards/coop.php -- Joseph Landman, Ph.D Founder and CEO Scalable Informatics LLC, email: landman@scalableinformatics.com web : http://www.scalableinformatics.com phone: +1 734 786 8423 fax : +1 734 786 8452 cell : +1 734 612 4615 From landman at scalableinformatics.com Sun Nov 5 16:05:46 2006 From: landman at scalableinformatics.com (Joe Landman) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <454E75F1.4010202@scalableinformatics.com> References: <454DED19.3060300@scalableinformatics.com> <005701c7012b$0a460150$9600000a@gourmandises> <454E75F1.4010202@scalableinformatics.com> Message-ID: <454E7C5A.6090409@scalableinformatics.com> Joe Landman wrote: >> http://www.mersenne.org/bench.htm > > Will pull it down, and see if I can get it going in under 10 minutes. It is hand coded 32 bit SSE2 assembler. Won't bother with it. Won't get much benefit on this platform apart from processor speed bumps. -- Joseph Landman, Ph.D Founder and CEO Scalable Informatics LLC, email: landman@scalableinformatics.com web : http://www.scalableinformatics.com phone: +1 734 786 8423 fax : +1 734 786 8452 cell : +1 734 612 4615 From jlb17 at duke.edu Mon Nov 6 03:57:14 2006 From: jlb17 at duke.edu (Joshua Baker-LePain) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <454DED19.3060300@scalableinformatics.com> References: <454DED19.3060300@scalableinformatics.com> Message-ID: On Sun, 5 Nov 2006 at 8:54am, Joe Landman wrote > Initial benchmarks (2.6 GHz clock) put it at about 17% faster than Opteron > 275 and Woodcrest 5150 (2.66 GHz) on a GAMESS test we use (1h41m for > Woodcrest and 275, and 1h26m for this unit). I'd be very interested to see any application benchmarks you've run. In my testing , Xeon 5140s blow the doors off Opteron 270s -- moreso than the clockspeed difference would indicate -- in almost all my benchmarks. I was leaning the Xeon way for any new purchases, but I haven't had a chance to test the new Opterons yet. -- Joshua Baker-LePain Department of Biomedical Engineering Duke University From laytonjb at charter.net Mon Nov 6 05:57:49 2006 From: laytonjb at charter.net (Jeffrey B. Layton) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: References: <454DED19.3060300@scalableinformatics.com> Message-ID: <454F3F5D.10709@charter.net> Joshua Baker-LePain wrote: > On Sun, 5 Nov 2006 at 8:54am, Joe Landman wrote > >> Initial benchmarks (2.6 GHz clock) put it at about 17% faster than >> Opteron 275 and Woodcrest 5150 (2.66 GHz) on a GAMESS test we use >> (1h41m for Woodcrest and 275, and 1h26m for this unit). > > I'd be very interested to see any application benchmarks you've run. > In my testing , Xeon 5140s > blow the doors off Opteron 270s -- moreso than the clockspeed > difference would indicate -- in almost all my benchmarks. I was > leaning the Xeon way for any new purchases, but I haven't had a chance > to test the new Opterons yet. It's fairly well known that Woodcrest is better than Opteron on LS-Dyna. The results at Topcrunch also support this. I haven't looked too closely at _why_ it's faster. Is it the bigger cache? Is it a better chip? I'm not sure. I'm also not sure if LS-Dyna is built with the Intel compilers or not. This will also have an impact on performance since you have to _dumb-down_ the compile options to get it to run on Opteron (I mentioned this to Joe Landman, but for everyone's edification, here's a link: http://www.swallowtail.org/naughty-intel.html that discusses how to fix the resulting Intel binaries to run better on Opteron or to patch the compiler itself). I haven't seen the Matlab results before. Those are interesting. Can you explain the benchmarks a bit more? Is it the stock Matlab or did you substitute the BLAS library? That might make a difference. Thanks! Jeff From rbw at ahpcrc.org Mon Nov 6 06:57:15 2006 From: rbw at ahpcrc.org (Richard Walsh) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <005701c7012b$0a460150$9600000a@gourmandises> References: <454DED19.3060300@scalableinformatics.com> <005701c7012b$0a460150$9600000a@gourmandises> Message-ID: <454F4D4B.50403@ahpcrc.org> Vincent Diepeveen wrote: > Impressive bandwidth. > > Is this the so called 'k8h' revision which has improved SSE2 throughput? The socket F series is generally referred to as K8L (revision H). The dual-core is already available and is primarily a pin-count and memory controller upgrade (to DDR2). Bus bandwidth is 10.67 GB/sec. The impoved widths in SSE and bus to L1 and L2 cache don't make it into the design until the quad core (Barcelona) late Q2 next year. This chip will have two memory controllers on die, although they will run at 6.4 GB/sec. each, rather than 10.67 each at first. Here is a very complete presentation on AMD's plans over the next year. http://www.tacc.utexas.edu/cluster2006/Herve_Chevanne.pdf In making purchasing decisions, I would consider the dual-core socket F especially knowing that it will be socket compatible with the quad-core chips released next year. I am having trouble imagining the performance on dual-socket, quad-core Intel systems all working through one Northbridge even through the Intel "Core" design is very good. AMD make a number of similar changes in the Barcelona, except it leaves the processor at onl 3-way super-scalar. rbw -- Richard B. Walsh "The world is given to me only once, not one existing and one perceived. The subject and object are but one." Erwin Schroedinger Project Manager Network Computing Services, Inc. Army High Performance Computing Research Center (AHPCRC) rbw@ahpcrc.org | 612.337.3467 ----------------------------------------------------------------------- This message (including any attachments) may contain proprietary or privileged information, the use and disclosure of which is legally restricted. If you have received this message in error please notify the sender by reply message, do not otherwise distribute it, and delete this message, with all of its contents, from your files. ----------------------------------------------------------------------- From rbw at ahpcrc.org Mon Nov 6 07:08:02 2006 From: rbw at ahpcrc.org (Richard Walsh) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <454DED19.3060300@scalableinformatics.com> References: <454DED19.3060300@scalableinformatics.com> Message-ID: <454F4FD2.30601@ahpcrc.org> Joe Landman wrote: > We are working on a machine with 2 Opteron 2218s. For laughs, I ran > streams on it. Here are 1,2,3,4 processor data. > > 1 processor: > > Copy: 5713.9944 0.0591 0.0560 0.0600 > Scale: 5713.9822 0.0587 0.0560 0.0600 > Add: 5454.2389 0.0911 0.0880 0.0920 > Triad: 5454.1576 0.0916 0.0880 0.0920 > Mmmm ... these actually seem low considering the bus bandwidth is supposed to be 10.67 GB/sec in the socket F. Using the 75% rule (1 silent read, 2 actual reads, and 1 write) from the triad shouldn't we get: 10.67 * .75 = ~8 GBs/sec > > 2 processors: > > Function Rate (MB/s) Avg time Min time Max time > Copy: 5713.8788 0.0591 0.0560 0.0640 > Scale: 5713.9822 0.0596 0.0560 0.0640 > Add: 5713.9538 0.0858 0.0840 0.0880 > Triad: 5713.9538 0.0871 0.0840 0.0880 > > > 3 processors: > > Copy: 8888.4000 0.0391 0.0360 0.0400 > Scale: 8888.4147 0.0391 0.0360 0.0400 > Add: 8570.9733 0.0573 0.0560 0.0600 > Triad: 8570.9551 0.0587 0.0560 0.0600 > > > 4 processors: > > Copy: 11428.1834 0.0293 0.0280 0.0320 > Scale: 11428.1347 0.0298 0.0280 0.0320 > Add: 10908.5812 0.0449 0.0440 0.0480 > Triad: 10908.6107 0.0440 0.0440 0.0440 Mmm ... similarly ... this seems to be only 50% of the full bandwidth of 2 socket F buses? > I did get one outlier at 4 CPUs. Could be a timing glitch. > > 4 processors: > Copy: 13332.7765 0.0302 0.0240 0.0360 > Scale: 13332.7765 0.0262 0.0240 0.0320 > Add: 13332.6220 0.0400 0.0360 0.0480 > Triad: 14999.0756 0.0391 0.0320 0.0480 These are the numbers I would expect here ... about 70% of 2 * 10.67 GBs/sec. I think that your out-"liar" is actually giving you the correct numbers. > > This is a dual socket 1207 motherboard. DDR2/533 ECC Registered > memory (will look back and check to be sure). > > Initial benchmarks (2.6 GHz clock) put it at about 17% faster than > Opteron 275 and Woodcrest 5150 (2.66 GHz) on a GAMESS test we use > (1h41m for Woodcrest and 275, and 1h26m for this unit). AMDs premium socket F is the 2220 SE which runs at 2.8 GHz. The new AMD 4 digit naming convention is discussed in the presentation I posted earlier. rbw -- Richard B. Walsh "The world is given to me only once, not one existing and one perceived. The subject and object are but one." Erwin Schroedinger Project Manager Network Computing Services, Inc. Army High Performance Computing Research Center (AHPCRC) rbw@ahpcrc.org | 612.337.3467 ----------------------------------------------------------------------- This message (including any attachments) may contain proprietary or privileged information, the use and disclosure of which is legally restricted. If you have received this message in error please notify the sender by reply message, do not otherwise distribute it, and delete this message, with all of its contents, from your files. ----------------------------------------------------------------------- From diep at xs4all.nl Mon Nov 6 07:18:18 2006 From: diep at xs4all.nl (Vincent Diepeveen) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Thought that this might be of interest References: <454DED19.3060300@scalableinformatics.com> <005701c7012b$0a460150$9600000a@gourmandises> <454F4D4B.50403@ahpcrc.org> Message-ID: <004701c701b6$cc9d1230$9600000a@gourmandises> Thanks for your info, this is very helpful. So until end 2007 the core2 annihilates any opteron system. Except of course when you're interested in just measuring bandwidth. So the K8L should then take over from core2 the performance reign again. Wasn't that K8L also going to do 4 instructions per cycle like core2 is doing? For my chess application of course more instructions per cycle means faster. Thanks, Vincent ----- Original Message ----- From: "Richard Walsh" To: "Vincent Diepeveen" ; "Beowulf Mailing List" Sent: Monday, November 06, 2006 3:57 PM Subject: Re: [Beowulf] Thought that this might be of interest > Vincent Diepeveen wrote: >> Impressive bandwidth. >> >> Is this the so called 'k8h' revision which has improved SSE2 throughput? > The socket F series is generally referred to as K8L (revision H). The > dual-core > is already available and is primarily a pin-count and memory > controller upgrade > (to DDR2). Bus bandwidth is 10.67 GB/sec. > > The impoved widths in SSE and bus to L1 and L2 cache don't make it > into the > design until the quad core (Barcelona) late Q2 next year. This chip > will have > two memory controllers on die, although they will run at 6.4 GB/sec. > each, rather > than 10.67 each at first. > > Here is a very complete presentation on AMD's plans over the next > year. > > http://www.tacc.utexas.edu/cluster2006/Herve_Chevanne.pdf > > In making purchasing decisions, I would consider the dual-core socket > F especially > knowing that it will be socket compatible with the quad-core chips > released next > year. I am having trouble imagining the performance on dual-socket, > quad-core > Intel systems all working through one Northbridge even through the > Intel "Core" > design is very good. AMD make a number of similar changes in the > Barcelona, > except it leaves the processor at onl 3-way super-scalar. > > rbw > > > > -- > > Richard B. Walsh > > "The world is given to me only once, not one existing and one > perceived. The subject and object are but one." > > Erwin Schroedinger > > Project Manager > Network Computing Services, Inc. > Army High Performance Computing Research Center (AHPCRC) > rbw@ahpcrc.org | 612.337.3467 > > ----------------------------------------------------------------------- > This message (including any attachments) may contain proprietary or > privileged information, the use and disclosure of which is legally > restricted. If you have received this message in error please notify > the sender by reply message, do not otherwise distribute it, and delete > this message, with all of its contents, from your files. > ----------------------------------------------------------------------- > > From jlb17 at duke.edu Mon Nov 6 07:29:44 2006 From: jlb17 at duke.edu (Joshua Baker-LePain) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <454F3F5D.10709@charter.net> References: <454DED19.3060300@scalableinformatics.com> <454F3F5D.10709@charter.net> Message-ID: On Mon, 6 Nov 2006 at 8:57am, Jeffrey B. Layton wrote > I'm also not sure if LS-Dyna is built with the Intel compilers or > not. This will also have an impact on performance since you have > to _dumb-down_ the compile options to get it to run on Opteron > (I mentioned this to Joe Landman, but for everyone's edification, > here's a link: They (LSTC) release binaries of each point revision compiled with both the PGI and the Intel compilers and name the binaries appropriately (I tested with both versions on both chips). For the MPI binaries, they make this explicit (see Table 3 in the PDF I linked to). For the SMP binaries, they call them 'amd' and 'xeon' (Tables 1 and 2). The README states that the compiler for the 'amd/PGI' versions is PGI 6.1 and for the 'xeon/Intel' version it's ifort 8.1. > I haven't seen the Matlab results before. Those are interesting. > Can you explain the benchmarks a bit more? Is it the stock Matlab I'll quote you what the grad student told me about the first one (Table 4), which was actually written by his undergrad: \begin{quote} It appears to be reading in FEM nodal data, then selecting a subset region to work with based upon user input. Nodal displacement data is (heavily) interpolated through time, then a subset time span is chosen. A (large) number of scatterers are then placed randomly inside the subset region. The delaunay transformation (computationally expensive) is used to find scatterers that would fall within the lumen of the heart chamber, then remove them. I don't know much about this step aside from the fact that it uses convex hulls. Once all the scatteres are in place, a trilinear interpolation is used to find the displacement of each scatterer based upon the motions of the nodal coordinates around it. \end{quote} Like I say in the table caption, that benchmark generates a large amount of system load -- it's about a 70/30 user/system split. Table 5 benchmarks Field II, which is an ultrasound simulation library for Matlab . It's distributed as a .mex file compiled *long* ago. So that benchmark is more about Field than about Matlab -- but it's relevant for us. > or did you substitute the BLAS library? That might make a > difference. That was stock Matlab R14p3. I'm in the process of getting an updated license file for the most recent version, and then maybe I'll look at playing with the BLAS library. -- Joshua Baker-LePain Department of Biomedical Engineering Duke University From rbw at ahpcrc.org Mon Nov 6 07:35:20 2006 From: rbw at ahpcrc.org (Richard Walsh) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <004701c701b6$cc9d1230$9600000a@gourmandises> References: <454DED19.3060300@scalableinformatics.com> <005701c7012b$0a460150$9600000a@gourmandises> <454F4D4B.50403@ahpcrc.org> <004701c701b6$cc9d1230$9600000a@gourmandises> Message-ID: <454F5638.7010009@ahpcrc.org> Vincent Diepeveen wrote: > Thanks for your info, this is very helpful. > > So until end 2007 the core2 annihilates any opteron system. Nope. Dual-core socket F does quite a bit to even the score on floating-point, just with the DDR2 latency and bandwidth improvements. I would not use the word annihilate ... but, Woodcrest still dominates for integer. > > Except of course when you're interested in just measuring bandwidth. > > So the K8L should then take over from core2 the performance reign again. While buying Woodcrest now might make sense on a raw performance basis, I would want to consider carefully the direction that AMD takes you in through 2007. The socket compatiblity of the the quad-core Barcelona is an important consideration especially considering what 8-cores on a board probably means for the Intel CloverTown. A key question is how well does your code use cache as always. > > Wasn't that K8L also going to do 4 instructions per cycle like core2 > is doing? That is not what I have read. See the presentation ... and other articles on Barcelona. I see only 3-way super scalar. > For my chess application of course more instructions per cycle means > faster. Yes. I can the that the bandwidth, instruction intensity, and integer profile of a chess application might favor the Woodcrest, on the other had a 4 socket socket 1207 motherboard with 8 channels to memory and 16 cores might be a better alternative as scale than an Intel SMP system. rbw -- Richard B. Walsh "The world is given to me only once, not one existing and one perceived. The subject and object are but one." Erwin Schroedinger Project Manager Network Computing Services, Inc. Army High Performance Computing Research Center (AHPCRC) rbw@ahpcrc.org | 612.337.3467 ----------------------------------------------------------------------- This message (including any attachments) may contain proprietary or privileged information, the use and disclosure of which is legally restricted. If you have received this message in error please notify the sender by reply message, do not otherwise distribute it, and delete this message, with all of its contents, from your files. ----------------------------------------------------------------------- From deadline at clustermonkey.net Mon Nov 6 07:45:37 2006 From: deadline at clustermonkey.net (Douglas Eadline) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] SC06 - shameless plug (sort of) Message-ID: <42646.192.168.1.1.1162827937.squirrel@mail.eadline.org> If you are heading to SC06 in Tampa next week, you may be interested in the following (if not just ignore this mailing). Over the last month or so I have been banging on two dual socket quad-core 1U nodes from Appro. That is 16 cores total. The nodes are connected by GigE and Mellanox Infiniband. I'll be in the booth at various times running some codes. If anyone wants to run a specific codes or just see what I have found (There is an white paper available) stop by booth. Bring your app on a memory stick. I'm interested in running different codes on these things. More information here: http://www.clustermonkey.net//content/view/169/2/ And don't forget the BeoBash: http://www.clustermonkey.net//content/view/166/2/ I'll be posting more info on ClusterMonkey during/after SC - can't release numbers until then. -- Doug From laytonjb at charter.net Mon Nov 6 07:54:44 2006 From: laytonjb at charter.net (Jeffrey B. Layton) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <004701c701b6$cc9d1230$9600000a@gourmandises> References: <454DED19.3060300@scalableinformatics.com> <005701c7012b$0a460150$9600000a@gourmandises> <454F4D4B.50403@ahpcrc.org> <004701c701b6$cc9d1230$9600000a@gourmandises> Message-ID: <454F5AC4.1000807@charter.net> OK, I feel like trying this one (call me stupid if you will). > Thanks for your info, this is very helpful. > > So until end 2007 the core2 annihilates any opteron system. No!!!!!!!!! This may be true for your application, but it's not true for all applications. I have a couple of applications where a 2.6 GHz Opteron with beat a 3.0 GHz Woodcrest. Another example is Fluent, where a 3.0 GHz Woodcrest kills the Opteron on a single node or a small number of nodes. But as you increase the number of nodes, the Opteron is faster than the Woodcrest. So which one is faster - Woodcrest or Opteron - really depends on your code. Jeff From landman at scalableinformatics.com Mon Nov 6 08:04:00 2006 From: landman at scalableinformatics.com (Joe Landman) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <454F4FD2.30601@ahpcrc.org> References: <454DED19.3060300@scalableinformatics.com> <454F4FD2.30601@ahpcrc.org> Message-ID: <454F5CF0.5080609@scalableinformatics.com> Richard Walsh wrote: > Joe Landman wrote: >> We are working on a machine with 2 Opteron 2218s. For laughs, I ran >> streams on it. Here are 1,2,3,4 processor data. >> >> 1 processor: >> >> Copy: 5713.9944 0.0591 0.0560 0.0600 >> Scale: 5713.9822 0.0587 0.0560 0.0600 >> Add: 5454.2389 0.0911 0.0880 0.0920 >> Triad: 5454.1576 0.0916 0.0880 0.0920 >> > Mmmm ... these actually seem low considering the bus bandwidth is > supposed to be 10.67 GB/sec Heh.... don't shoot the measurer ... :) I agree BTW that I expected this to be about 7++ GB/s per socket. This is a pathscale built binary. Will try with PGI and Intel as well. > in the socket F. Using the 75% rule (1 silent read, 2 actual reads, > and 1 write) from the triad shouldn't > we get: > > 10.67 * .75 = ~8 GBs/sec Possibly... The bandwidth per socket is 10.67 GB/s. So the two thread on one socket should be pretty darned close to that. I am surprised that it wasn't. Moreover, other folks seem to note what we observed: http://tweakers.net/reviews/646/5 with about 5.3 GB/s BW/core. I think affinity is keeping two threads on the same core, so it might take a little bit of work to get a real 1 thread/socket number. [...] >> 4 processors: >> Copy: 13332.7765 0.0302 0.0240 0.0360 >> Scale: 13332.7765 0.0262 0.0240 0.0320 >> Add: 13332.6220 0.0400 0.0360 0.0480 >> Triad: 14999.0756 0.0391 0.0320 0.0480 > These are the numbers I would expect here ... about 70% of 2 * > 10.67 GBs/sec. > I think that your out-"liar" is actually giving you the correct > numbers. Possible. I was expecting something north of 12 GB/s. -- Joseph Landman, Ph.D Founder and CEO Scalable Informatics LLC, email: landman@scalableinformatics.com web : http://www.scalableinformatics.com phone: +1 734 786 8423 fax : +1 734 786 8452 or +1 866 888 3112 cell : +1 734 612 4615 From landman at scalableinformatics.com Mon Nov 6 08:33:35 2006 From: landman at scalableinformatics.com (Joe Landman) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <454F5AC4.1000807@charter.net> References: <454DED19.3060300@scalableinformatics.com> <005701c7012b$0a460150$9600000a@gourmandises> <454F4D4B.50403@ahpcrc.org> <004701c701b6$cc9d1230$9600000a@gourmandises> <454F5AC4.1000807@charter.net> Message-ID: <454F63DF.1090803@scalableinformatics.com> Jeffrey B. Layton wrote: [...] > So which one is faster - Woodcrest or Opteron - really > depends on your code. Yup. Most of the testing I have done, where Woodcrest is faster appears to be linked to the cache size. Since I used the PGI compilers, and built a single binary (not the mixed binary), it sort of pulls this out and highlights it. A CPU with 4M apparent cache for a single processor will run circles around a machine with a 1M apparent cache per processor for cache friendly apps running a single thread. As you increase the number of threads per socket, things even out. And this jives well with my testing to date. That said, I am quite interested in trying to feed the Woodcrest SSE pipeline at a better rate than the compilers allow for (same with the Opteron). Unfortunately this gets into the area of hand coded SSE*, and the Intel compiler does a decent job supporting this (while the PGI doesn't). Since the Intel compiler has other things working against it (ahem!), this makes the job of writing fast code somewhat harder. I know, lets all just use GCC ... -- Joseph Landman, Ph.D Founder and CEO Scalable Informatics LLC, email: landman@scalableinformatics.com web : http://www.scalableinformatics.com phone: +1 734 786 8423 fax : +1 734 786 8452 or +1 866 888 3112 cell : +1 734 612 4615 From ctierney at hypermall.net Mon Nov 6 08:42:42 2006 From: ctierney at hypermall.net (Craig Tierney) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <454F5AC4.1000807@charter.net> References: <454DED19.3060300@scalableinformatics.com> <005701c7012b$0a460150$9600000a@gourmandises> <454F4D4B.50403@ahpcrc.org> <004701c701b6$cc9d1230$9600000a@gourmandises> <454F5AC4.1000807@charter.net> Message-ID: <454F6602.5000203@hypermall.net> Jeffrey B. Layton wrote: > OK, I feel like trying this one (call me stupid if you will). > >> Thanks for your info, this is very helpful. >> >> So until end 2007 the core2 annihilates any opteron system. > > No!!!!!!!!! This may be true for your application, but it's not > true for all applications. I have a couple of applications where > a 2.6 GHz Opteron with beat a 3.0 GHz Woodcrest. > > Another example is Fluent, where a 3.0 GHz Woodcrest kills > the Opteron on a single node or a small number of nodes. > But as you increase the number of nodes, the Opteron is faster > than the Woodcrest. Why is this? Is it the interconnect, PCI-E implementation, maybe cache effects? Craig > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > From twhigham1 at yahoo.com Wed Nov 1 14:19:15 2006 From: twhigham1 at yahoo.com (Tom Whigham) Date: Fri May 9 01:05:26 2008 Subject: [Beowulf] Beginners Forum info! Message-ID: <20061101221915.57018.qmail@web52112.mail.yahoo.com> Is there a beginners forum to subscribe to? I just purchased 4 used PIII computers and Red Hat 7.1 and would like to begin learning about cluster computing. Where should I start? ____________________________________________________________________________________ Get your email and see which of your friends are online - Right on the New Yahoo.com (http://www.yahoo.com/preview) From shyaam at gmail.com Mon Nov 6 06:31:35 2006 From: shyaam at gmail.com (Shyaam) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <454DED19.3060300@scalableinformatics.com> References: <454DED19.3060300@scalableinformatics.com> Message-ID: Is Simple Scalar(http://simplescalar.com/) good forr testing beowulf clusters? Kind regards, Shyaam -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.scyld.com/pipermail/beowulf/attachments/20061106/f2453776/attachment.html From diep at xs4all.nl Mon Nov 6 08:46:12 2006 From: diep at xs4all.nl (Vincent Diepeveen) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest References: <454DED19.3060300@scalableinformatics.com> <005701c7012b$0a460150$9600000a@gourmandises> <454F4D4B.50403@ahpcrc.org> <004701c701b6$cc9d1230$9600000a@gourmandises> <454F5638.7010009@ahpcrc.org> Message-ID: <005c01c701c3$1e78eb40$9600000a@gourmandises> Thanks for your answer. Please show us results of prime95 (iteration times is already enough) please to prove your marketing talk regarding floating point unit of existing K8 chips.. Paper benchmarks of items that are not even close to be able to get bought in shops happen just a bit too much IMHO. Even an old P4 2.4Ghz is faster than my dual core opteron 2.4Ghz here for prime95 and LLR. Core2 at single core is 2 times faster than that P4 at 2.4Ghz. RAM latency is not that interesting for most SIMD applications such as prime95, instead it is more interesting for integer applications to have a good latency, but even then we're talking about 5%-10% of the system time at most. Further core2 has a 2 times bigger L2 cache than K8, so that hides quite some latency. Last but not least you should also theoretically go in a room with no contact to the rest of the world and ponder about next statement: "Imagine you are a programmer and your number crunching applications scaling is totally dependant upon bandwidth; shouldn't you take a different job than programming?" Vincent ----- Original Message ----- From: "Richard Walsh" To: "Vincent Diepeveen" Cc: "Beowulf Mailing List" Sent: Monday, November 06, 2006 4:35 PM Subject: Re: [Beowulf] Thought that this might be of interest > Vincent Diepeveen wrote: >> Thanks for your info, this is very helpful. >> >> So until end 2007 the core2 annihilates any opteron system. > Nope. Dual-core socket F does quite a bit to even the score > on floating-point, just with the DDR2 latency and bandwidth > improvements. I would not use the word annihilate ... but, > Woodcrest still dominates for integer. >> >> Except of course when you're interested in just measuring bandwidth. >> >> So the K8L should then take over from core2 the performance reign again. > While buying Woodcrest now might make sense on a raw performance > basis, I would want to consider carefully the direction that AMD takes > you in through 2007. The socket compatiblity of the the quad-core > Barcelona > is an important consideration especially considering what 8-cores on a > board probably means for the Intel CloverTown. A key question is how > well does your code use cache as always. >> >> Wasn't that K8L also going to do 4 instructions per cycle like core2 is >> doing? > That is not what I have read. See the presentation ... and other > articles on > Barcelona. I see only 3-way super scalar. >> For my chess application of course more instructions per cycle means >> faster. > Yes. I can the that the bandwidth, instruction intensity, and integer > profile of > a chess application might favor the Woodcrest, on the other had a 4 > socket > socket 1207 motherboard with 8 channels to memory and 16 cores might be > a better alternative as scale than an Intel SMP system. > > rbw > > -- > > Richard B. Walsh > > "The world is given to me only once, not one existing and one > perceived. The subject and object are but one." > > Erwin Schroedinger > > Project Manager > Network Computing Services, Inc. > Army High Performance Computing Research Center (AHPCRC) > rbw@ahpcrc.org | 612.337.3467 > > ----------------------------------------------------------------------- > This message (including any attachments) may contain proprietary or > privileged information, the use and disclosure of which is legally > restricted. If you have received this message in error please notify > the sender by reply message, do not otherwise distribute it, and delete > this message, with all of its contents, from your files. > ----------------------------------------------------------------------- > > From deadline at eadline.org Mon Nov 6 09:08:05 2006 From: deadline at eadline.org (Douglas Eadline) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Beginners Forum info! In-Reply-To: <20061101221915.57018.qmail@web52112.mail.yahoo.com> References: <20061101221915.57018.qmail@web52112.mail.yahoo.com> Message-ID: <53438.192.168.1.1.1162832885.squirrel@mail.eadline.org> Check out ClusterMonkey.net Particularly the New to Clusters section. It will help you navigate some of the issues. http://www.clustermonkey.net//content/view/91/44/ -- Doug > Is there a beginners forum to subscribe to? > I just purchased 4 used PIII computers and Red Hat 7.1 > and would like to begin learning about cluster > computing. Where should I start? > > > > ____________________________________________________________________________________ > Get your email and see which of your friends are online - Right on the New > Yahoo.com > (http://www.yahoo.com/preview) > > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > -- Doug From greg.lindahl at qlogic.com Mon Nov 6 10:27:37 2006 From: greg.lindahl at qlogic.com (Greg Lindahl) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <454E75F1.4010202@scalableinformatics.com> References: <454DED19.3060300@scalableinformatics.com> <005701c7012b$0a460150$9600000a@gourmandises> <454E75F1.4010202@scalableinformatics.com> Message-ID: <20061106182737.GH1892@greglaptop.rchland.ibm.com> On Sun, Nov 05, 2006 at 06:38:25PM -0500, Joe Landman wrote: > Since they wish to do it only for Intel processors, and the world > is decidedly mixed, this has implications on the use of Intel compilers > for lots of people wishing to get the best performance on all platforms > with a single compiler tool. Doesn't it. I will note that on Woodcrest & Core2 cores, the PathScale compilers do (relatively speaking) a lot better than on previous Intel cpus. So there's an option. Also, anyone benchmarking Woodcrest vs. Opteron should be sure to run on all the cores in the system, not just one, unless your real production runs only use 1 core/node. One common situation with Woodcrest is that the single-core performance beats everyone else, but scaling to using all the cores in your node is less than the competition. So the final answer is extremely application dependent. Some codes are also outliers, for example Fluent on Woodcrest does great, if I remember correctly. -- greg From laytonjb at charter.net Mon Nov 6 10:41:15 2006 From: laytonjb at charter.net (Jeffrey B. Layton) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <20061106182737.GH1892@greglaptop.rchland.ibm.com> References: <454DED19.3060300@scalableinformatics.com> <005701c7012b$0a460150$9600000a@gourmandises> <454E75F1.4010202@scalableinformatics.com> <20061106182737.GH1892@greglaptop.rchland.ibm.com> Message-ID: <454F81CB.1030806@charter.net> Greg Lindahl wrote: > On Sun, Nov 05, 2006 at 06:38:25PM -0500, Joe Landman wrote: > > >> Since they wish to do it only for Intel processors, and the world >> is decidedly mixed, this has implications on the use of Intel compilers >> for lots of people wishing to get the best performance on all platforms >> with a single compiler tool. Doesn't it. >> > > > Some codes are also outliers, for example Fluent on Woodcrest does > great, if I remember correctly. > The Fluent benchmarks need to be explained. There are basically 9 benchmarks (small, medium, and large models). On the small and medium models the Woodcrest kicks butt on a single node (1-4 cores) mostly due to the cache size. On some of the large models, the Woodcrest does very well on a single node (1-4 CPUs). But on some of the large models, the Opteron does very well. The small and medium models are really pretty small so when I look at more than 1 node, I look at the large models. At a certain point, the slower Opterons become faster than Woodcrest (both IB). I also know one CFD app that is faster on Opteron even for one node (1-4 CPUs). Jeff From atchley at myri.com Mon Nov 6 12:09:59 2006 From: atchley at myri.com (Scott Atchley) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <454F81CB.1030806@charter.net> References: <454DED19.3060300@scalableinformatics.com> <005701c7012b$0a460150$9600000a@gourmandises> <454E75F1.4010202@scalableinformatics.com> <20061106182737.GH1892@greglaptop.rchland.ibm.com> <454F81CB.1030806@charter.net> Message-ID: <10FDA19E-A638-49B9-AB2B-037F651C2C70@myri.com> On Nov 6, 2006, at 1:41 PM, Jeffrey B. Layton wrote: >> Some codes are also outliers, for example Fluent on Woodcrest does >> great, if I remember correctly. > > The Fluent benchmarks need to be explained. There are basically > 9 benchmarks (small, medium, and large models). On the small > and medium models the Woodcrest kicks butt on a single node > (1-4 cores) mostly due to the cache size. On some of the large > models, the Woodcrest does very well on a single node (1-4 CPUs). > But on some of the large models, the Opteron does very well. While benchmarking MX-10G, my contacts at Fluent said that only the three large models are useful these days for comparing systems (i.e. the small and medium datasets no longer represent what their customers use). Scott -- Scott Atchley Myricom Inc. http://www.myri.com From rbw at ahpcrc.org Mon Nov 6 12:19:48 2006 From: rbw at ahpcrc.org (Richard Walsh) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <454F5CF0.5080609@scalableinformatics.com> References: <454DED19.3060300@scalableinformatics.com> <454F4FD2.30601@ahpcrc.org> <454F5CF0.5080609@scalableinformatics.com> Message-ID: <454F98E4.1010203@ahpcrc.org> Joe Landman wrote: > Richard Walsh wrote: >> Joe Landman wrote: >>> We are working on a machine with 2 Opteron 2218s. For laughs, I ran >>> streams on it. Here are 1,2,3,4 processor data. >>> >>> 1 processor: >>> >>> Copy: 5713.9944 0.0591 0.0560 0.0600 >>> Scale: 5713.9822 0.0587 0.0560 0.0600 >>> Add: 5454.2389 0.0911 0.0880 0.0920 >>> Triad: 5454.1576 0.0916 0.0880 0.0920 >>> >> Mmmm ... these actually seem low considering the bus bandwidth is >> supposed to be 10.67 GB/sec > > Heh.... don't shoot the measurer ... :) Never measurers are honored guests at my table ... ;-) > > I agree BTW that I expected this to be about 7++ GB/s per socket. > This is a pathscale built binary. Will try with PGI and Intel as well. In the presentation I directed folks to there are socket F stream numbers for both 2x2 and 4x2 cases (4 and 8 threads). These are AMDs numbers, but they report: 2x2 4 threads Intel 5160 3.0 GHz at 5865 for the triad (27% of maximum) 2x2 4 threads Opteron 2220 SE 2.8 GHz at 13296 for the triad (62 % of maximum) 4x2 8 threads Opteron 2220 SE 2.8 GHz at 18271 for the triad (42 % of maximum) These numbers are both good and bad. Good in the sense that socket F shows fairly good percentages of designed bandwidth. Bad in the sense that one might expect--less and less of it seems usable as the core/thread count go up. A 2x4 system I would expect to do better than an 4x2 system on a percentage basis, but the channels are a bit slower on Barcelona (back at socket 940 6.4 GBs levels), so the absolute number may be the same or worse. rbw -- Richard B. Walsh "The world is given to me only once, not one existing and one perceived. The subject and object are but one." Erwin Schroedinger Project Manager Network Computing Services, Inc. Army High Performance Computing Research Center (AHPCRC) rbw@ahpcrc.org | 612.337.3467 ----------------------------------------------------------------------- This message (including any attachments) may contain proprietary or privileged information, the use and disclosure of which is legally restricted. If you have received this message in error please notify the sender by reply message, do not otherwise distribute it, and delete this message, with all of its contents, from your files. ----------------------------------------------------------------------- From hahn at physics.mcmaster.ca Mon Nov 6 12:42:24 2006 From: hahn at physics.mcmaster.ca (Mark Hahn) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: References: <454DED19.3060300@scalableinformatics.com> Message-ID: > Is Simple Scalar(http://simplescalar.com/) good forr testing beowulf > clusters? but why? simplescalar appears to be an ISA simulator, which I would guess would exibit a relatively small memory footprint and lots of branchy, mostly-int code. if that's the case, then it would make a pretty poor cluster tester. HPL is a pretty decent way to test a cluster, since it can be configured to fill your memory, will certainly keep your FPU warm, as well as make realistic (if not extreme) use of your interconnect. regards, mark hahn. From hahn at physics.mcmaster.ca Mon Nov 6 12:45:58 2006 From: hahn at physics.mcmaster.ca (Mark Hahn) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Beginners Forum info! In-Reply-To: <20061101221915.57018.qmail@web52112.mail.yahoo.com> References: <20061101221915.57018.qmail@web52112.mail.yahoo.com> Message-ID: > Is there a beginners forum to subscribe to? why a forum? there are tons of excellent resources on the web that you can read to get up-to-speed. it's a lot more time-effective for everyone if you pick up the background first. > I just purchased 4 used PIII computers and Red Hat 7.1 why the heck would you ever bother with RH 7.1?!? > and would like to begin learning about cluster > computing. Where should I start? for a demo cluster like this, just install packages from some decent distro (FC, for instance, would support your hardware and comes with preconfigured packages for mpich/lam, etc.) regards, mark hahn. From 3lucid at gmail.com Mon Nov 6 14:23:20 2006 From: 3lucid at gmail.com (Kyle Spaans) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Beginners Forum info! In-Reply-To: <20061101221915.57018.qmail@web52112.mail.yahoo.com> References: <20061101221915.57018.qmail@web52112.mail.yahoo.com> Message-ID: <5a1205b30611061423l2bd690d3u201f21d3e4fa8926@mail.gmail.com> Welcome! You're in the same boat as me. I've found that just browsing ClusterMonkey, and reading the intro document on the beowulf.org website to be very helpful. Being subscribed to this list also helps. The odd relevant [to us begginners that is] discussion.If you are looking for things to read, just browse the mailling list archives, looking for keywords. For conversation sake, are you also 17 and also starting University? gl & hf From James.P.Lux at jpl.nasa.gov Mon Nov 6 14:38:41 2006 From: James.P.Lux at jpl.nasa.gov (Jim Lux) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Beginners Forum info! In-Reply-To: References: <20061101221915.57018.qmail@web52112.mail.yahoo.com> Message-ID: <6.2.3.4.2.20061106143557.0322dad8@mail.jpl.nasa.gov> At 12:45 PM 11/6/2006, Mark Hahn wrote: >>Is there a beginners forum to subscribe to? > >why a forum? there are tons of excellent resources on the web >that you can read to get up-to-speed. it's a lot more time-effective >for everyone if you pick up the background first. > >>I just purchased 4 used PIII computers and Red Hat 7.1 > >why the heck would you ever bother with RH 7.1?!? > >>and would like to begin learning about cluster >>computing. Where should I start? You could use something like Cluster Knoppix. Burn 4 copies of the disk, fling them into the drives, fire up the computers, and go to town. But another poster's suggestion go to ClusterMonkey is a good one. There's some tutorials there on getting revved up on the first go around. ClusterWorld magazine had some tutorials too (is it online? I just have the paper copies lying around) Burn a few disks, fire up that MPI POVray, etc. James Lux, P.E. Spacecraft Radio Frequency Subsystems Group Flight Communications Systems Section Jet Propulsion Laboratory, Mail Stop 161-213 4800 Oak Grove Drive Pasadena CA 91109 tel: (818)354-2075 fax: (818)393-6875 From deadline at clustermonkey.net Mon Nov 6 15:44:04 2006 From: deadline at clustermonkey.net (Douglas Eadline) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Beginners Forum info! In-Reply-To: <5a1205b30611061423l2bd690d3u201f21d3e4fa8926@mail.gmail.com> References: <20061101221915.57018.qmail@web52112.mail.yahoo.com> <5a1205b30611061423l2bd690d3u201f21d3e4fa8926@mail.gmail.com> Message-ID: <48022.192.168.1.1.1162856644.squirrel@mail.eadline.org> Future Cluster Geeks: I would be very interested if there is anything that you think is "missing" in your search. i.e. what big questions do you have to which you cannot easily find an answer ? My invitation extends to all the beginners on the list. No question is to stupid or silly. Reading these posts I think a "Beginning Cluster FAQ" would be a nice addition to ClusterMonkey. Speaking for myself, and I assume many others on this list, I am too far entrenched in this stuff to see the nagging issues and questions that beginners face. Please feel free to drop me a line or post to the list. -- Doug > Welcome! > You're in the same boat as me. > I've found that just browsing ClusterMonkey, and reading the intro > document on the beowulf.org website to be very helpful. Being > subscribed to this list also helps. The odd relevant [to us begginners > that is] discussion.If you are looking for things to read, just browse > the mailling list archives, looking for keywords. > For conversation sake, are you also 17 and also starting University? > > gl & hf > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > -- Doug From deadline at clustermonkey.net Mon Nov 6 15:47:37 2006 From: deadline at clustermonkey.net (Douglas Eadline) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Beginners Forum info! In-Reply-To: <6.2.3.4.2.20061106143557.0322dad8@mail.jpl.nasa.gov> References: <20061101221915.57018.qmail@web52112.mail.yahoo.com> <6.2.3.4.2.20061106143557.0322dad8@mail.jpl.nasa.gov> Message-ID: <45411.192.168.1.1.1162856857.squirrel@mail.eadline.org> --snip-- > But another poster's suggestion go to ClusterMonkey is a good > one. There's some tutorials there on getting revved up on the first > go around. ClusterWorld magazine had some tutorials too (is it > online? I just have the paper copies lying around) Actually ClusterMonkey is running some of the old ClusterWorld content. ClusterWorld never made it to the web, but the relevant content, with owners permission, will probably show up on ClusterMonkey at some point. -- Doug > > Burn a few disks, fire up that MPI POVray, etc. > > > > > > James Lux, P.E. > Spacecraft Radio Frequency Subsystems Group > Flight Communications Systems Section > Jet Propulsion Laboratory, Mail Stop 161-213 > 4800 Oak Grove Drive > Pasadena CA 91109 > tel: (818)354-2075 > fax: (818)393-6875 > > > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > -- Doug From James.P.Lux at jpl.nasa.gov Mon Nov 6 17:12:03 2006 From: James.P.Lux at jpl.nasa.gov (Jim Lux) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Beginners Forum info! In-Reply-To: <5a1205b30611061423l2bd690d3u201f21d3e4fa8926@mail.gmail.co m> References: <20061101221915.57018.qmail@web52112.mail.yahoo.com> <5a1205b30611061423l2bd690d3u201f21d3e4fa8926@mail.gmail.com> Message-ID: <6.2.3.4.2.20061106171039.03218f08@mail.jpl.nasa.gov> At 02:23 PM 11/6/2006, Kyle Spaans wrote: >Welcome! >You're in the same boat as me. >I've found that just browsing ClusterMonkey, and reading the intro >document on the beowulf.org website to be very helpful. Being >subscribed to this list also helps. The odd relevant [to us begginners >that is] discussion.If you are looking for things to read, just browse >the mailling list archives, looking for keywords. >For conversation sake, are you also 17 and also starting University? Robert G Brown at Duke (rgb on the list) has a good online book on beowulfery at his website: http://www.phy.duke.edu/~rgb/ >gl & hf >_______________________________________________ >Beowulf mailing list, Beowulf@beowulf.org >To change your subscription (digest mode or unsubscribe) visit >http://www.beowulf.org/mailman/listinfo/beowulf James Lux, P.E. Spacecraft Radio Frequency Subsystems Group Flight Communications Systems Section Jet Propulsion Laboratory, Mail Stop 161-213 4800 Oak Grove Drive Pasadena CA 91109 tel: (818)354-2075 fax: (818)393-6875 From csamuel at vpac.org Mon Nov 6 18:29:03 2006 From: csamuel at vpac.org (Chris Samuel) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <004701c701b6$cc9d1230$9600000a@gourmandises> References: <454DED19.3060300@scalableinformatics.com> <454F4D4B.50403@ahpcrc.org> <004701c701b6$cc9d1230$9600000a@gourmandises> Message-ID: <200611071329.03296.csamuel@vpac.org> On Tuesday 07 November 2006 02:18, Vincent Diepeveen wrote: > So until end 2007 the core2 annihilates any opteron system. In my experience with both architectures it depends on what you want out of a box, and what you're running on it, as well as your power constraints.. There is a good reason that the Apple Woodcrest boxes I've seen have heat sinks on their FB-DIMMs.. -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia From csamuel at vpac.org Mon Nov 6 18:34:25 2006 From: csamuel at vpac.org (Chris Samuel) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <20061106182737.GH1892@greglaptop.rchland.ibm.com> References: <454DED19.3060300@scalableinformatics.com> <454E75F1.4010202@scalableinformatics.com> <20061106182737.GH1892@greglaptop.rchland.ibm.com> Message-ID: <200611071334.25659.csamuel@vpac.org> On Tuesday 07 November 2006 05:27, Greg Lindahl wrote: > I will note that on Woodcrest & Core2 cores, the PathScale compilers > do (relatively speaking) a lot better than on previous Intel cpus. > So there's an option. Have they moved to FlexLM yet ? We had PathScale on eval for a while but found that their bespoke license daemon hung onto a users license for a long while after they'd finished compiling whereas the Intel compilers just worked.. cheers, Chris -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia From 3lucid at gmail.com Mon Nov 6 20:13:18 2006 From: 3lucid at gmail.com (Kyle Spaans) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Beginners Forum info! In-Reply-To: <6.2.3.4.2.20061106171039.03218f08@mail.jpl.nasa.gov> References: <20061101221915.57018.qmail@web52112.mail.yahoo.com> <5a1205b30611061423l2bd690d3u201f21d3e4fa8926@mail.gmail.com> <6.2.3.4.2.20061106171039.03218f08@mail.jpl.nasa.gov> Message-ID: <5a1205b30611062013u42ba1859hc2df7bf36d2b5d3d@mail.gmail.com> Actually some keywords I forgot to mention that might come in handy are: Scyld, Rocks, Oscar, Condor From steve_heaton at iinet.net.au Mon Nov 6 20:39:31 2006 From: steve_heaton at iinet.net.au (steve_heaton@iinet.net.au) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Beginners Forum info! Message-ID: <1162874371.45500e033adfc@mail.iinet.net.au> > I just purchased 4 used PIII computers and Red Hat 7.1 > and would like to begin learning about cluster > computing. Where should I start? I think you'll get more out of it if you have a *reason* to run a Beowulf. Check the ClusterMonkey 'Value Cluster' artilces for a lot of useful ideas re suggestions for what do do with a cluster and actually going about building too: www.clustermonkey.net/content/view/41/33/ I also agree with Mark, go for a more recent distro. Plenty of options. There's your first round of questions: Which one? What difference does it make? Pros and cons. I was in a similar position about 3 years ago but I knew what I wanted. I was looking for a way to run galaxy smashes (LSS dynamics). I started with a P2-400, a P3-500 and a P4-2.8 all running FC3... ahhh those were the days. RSH or SSH? Surely SSH was slower... How do you get avoid a password prompt? Diskless or local disk? If local, how best to image? What sort of NIC? Do I need Jumbo packets? Do compiler options make any diff? Does the compiler? *Where* is all that wall clock time going? These and other questions. You'll soon learn the answer is usually 'it depends...' You just need to know what it depends on and the option that fits your needs the best. Welcome aboard :) Cheers Stevo From gdjacobs at gmail.com Mon Nov 6 20:43:22 2006 From: gdjacobs at gmail.com (Geoff Jacobs) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Beginners Forum info! In-Reply-To: <5a1205b30611062013u42ba1859hc2df7bf36d2b5d3d@mail.gmail.com> References: <20061101221915.57018.qmail@web52112.mail.yahoo.com> <5a1205b30611061423l2bd690d3u201f21d3e4fa8926@mail.gmail.com> <6.2.3.4.2.20061106171039.03218f08@mail.jpl.nasa.gov> <5a1205b30611062013u42ba1859hc2df7bf36d2b5d3d@mail.gmail.com> Message-ID: <45500EEA.4020608@gmail.com> Kyle Spaans wrote: > Actually some keywords I forgot to mention that might come in handy are: > Scyld, Rocks, Oscar, Condor Warewulf -- Geoffrey D. Jacobs Go to the Chinese Restaurant, Order the Special From deadline at clustermonkey.net Tue Nov 7 07:27:02 2006 From: deadline at clustermonkey.net (Douglas Eadline) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] LECCIBG Announcement (SC06) Message-ID: <56148.192.168.1.1.1162913222.squirrel@mail.eadline.org> For those attending SC06 and who want to make there cloths smell bad. The SC06 LECCIBG will be at the King Corona Cigars Bar/Club located at 1523 E 7th Ave Ybor City. The corner of N 15th and E 7th. We will start about 9:30 PM or so after the opening Gayla. More here: http://www.clustermonkey.net//content/view/164/57/ This is heart of the cigar world. Walt Ligon and I have been training six years for this. -- Doug From diep at xs4all.nl Tue Nov 7 09:17:29 2006 From: diep at xs4all.nl (Vincent Diepeveen) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest References: <454DED19.3060300@scalableinformatics.com> <005701c7012b$0a460150$9600000a@gourmandises> <454E75F1.4010202@scalableinformatics.com><20061106182737.GH1892@greglaptop.rchland.ibm.com> <454F81CB.1030806@charter.net> Message-ID: <005901c70290$9a9dd690$9600000a@gourmandises> In short a lot of confusion and that all for nothing. In those CFD codes probably more RAM is better. You prefer a 0.8Ghz processor with 1 terabyte of RAM over a 8Ghz Core2 with 4GB ram. The usage of the word 'node' is total out of context here. We speak only about a machine with a lot of RAM and in fact i've been using a few processors from itanium systems where i used a few MB of RAM fitting in L3 per core (for sieving), meanwhile one guy running CFD just wanted to use all the RAM just using 1 core for the 'calculations'. That was a good usage of the hardware! CPU speed is total irrelevant here. Just give them a terabyte of DDR2 RAM i'd say. Therefore Jeffrey, buying new cpu's for this is total irrelevant, you probably prefer a single core 2.8Ghz A64 with a LOT of DDR2 ram for this and give all those other cores to people who can use cpu speed. Note you might want to consider other solutions for CFD codes. There is so 'ram harddrives' with very fast access times to something that is seemingly a harddrive but in fact just a big bunch of RAM. the word 'node' is total irrelevant in this discussion. It just goes about RAM size and only long after that the latency to the RAM. ----- Original Message ----- From: "Jeffrey B. Layton" To: "beowulf" Sent: Monday, November 06, 2006 7:41 PM Subject: Re: [Beowulf] Thought that this might be of interest > Greg Lindahl wrote: >> On Sun, Nov 05, 2006 at 06:38:25PM -0500, Joe Landman wrote: >> >> >>> Since they wish to do it only for Intel processors, and the world is >>> decidedly mixed, this has implications on the use of Intel compilers for >>> lots of people wishing to get the best performance on all platforms with >>> a single compiler tool. Doesn't it. >>> >> >> >> Some codes are also outliers, for example Fluent on Woodcrest does >> great, if I remember correctly. >> > > The Fluent benchmarks need to be explained. There are basically > 9 benchmarks (small, medium, and large models). On the small > and medium models the Woodcrest kicks butt on a single node > (1-4 cores) mostly due to the cache size. On some of the large > models, the Woodcrest does very well on a single node (1-4 CPUs). > But on some of the large models, the Opteron does very well. > > The small and medium models are really pretty small so when > I look at more than 1 node, I look at the large models. At a > certain point, the slower Opterons become faster than Woodcrest > (both IB). > > I also know one CFD app that is faster on Opteron even for one > node (1-4 CPUs). > > Jeff > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > From wrankin at ee.duke.edu Tue Nov 7 11:23:31 2006 From: wrankin at ee.duke.edu (Bill Rankin) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] LECCIBG Announcement (SC06) In-Reply-To: <56148.192.168.1.1.1162913222.squirrel@mail.eadline.org> References: <56148.192.168.1.1.1162913222.squirrel@mail.eadline.org> Message-ID: <771A122E-AC75-49C2-876F-F18827CB034E@ee.duke.edu> On Nov 7, 2006, at 10:27 AM, Douglas Eadline wrote: > This is heart of the cigar world. Walt Ligon and I have been > training six > years for this. > Well Doug, for us relative cigar novices, what advice would you offer as to preparation for this event? I assume that there will be opportunity for procurement of cigars for the event? I am already aware of the need for wearing sacrificial clothing ;-) -b From john.leidel at gmail.com Tue Nov 7 12:01:52 2006 From: john.leidel at gmail.com (John Leidel) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] LECCIBG Announcement (SC06) In-Reply-To: <771A122E-AC75-49C2-876F-F18827CB034E@ee.duke.edu> References: <56148.192.168.1.1.1162913222.squirrel@mail.eadline.org> <771A122E-AC75-49C2-876F-F18827CB034E@ee.duke.edu> Message-ID: <1162929712.5766.5.camel@athlon64.site> I would certainly suggest making a stop at the Thompsons Cigar factory store on the way from the airport. http://maps.google.com/maps?f=q&hl=en&q=thompsons+cigars+near+Tampa, +FL&ie=UTF8&z=13&ll=27.989854,-82.479515&spn=0.104895,0.161018&om=1&iwloc=A :-) cheers john On Tue, 2006-11-07 at 14:23 -0500, Bill Rankin wrote: > On Nov 7, 2006, at 10:27 AM, Douglas Eadline wrote: > > > This is heart of the cigar world. Walt Ligon and I have been > > training six > > years for this. > > > > Well Doug, for us relative cigar novices, what advice would you offer > as to preparation for this event? I assume that there will be > opportunity for procurement of cigars for the event? > > I am already aware of the need for wearing sacrificial clothing ;-) > > -b > > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf From deadline at clustermonkey.net Tue Nov 7 12:37:27 2006 From: deadline at clustermonkey.net (Douglas Eadline) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] LECCIBG Announcement (SC06) In-Reply-To: <771A122E-AC75-49C2-876F-F18827CB034E@ee.duke.edu> References: <56148.192.168.1.1.1162913222.squirrel@mail.eadline.org> <771A122E-AC75-49C2-876F-F18827CB034E@ee.duke.edu> Message-ID: <40975.192.168.1.1.1162931847.squirrel@mail.eadline.org> Bill, First, for the purpose of political correctness: Cigars can be harmful to your health, they can stink (cheap ones do), one end is hot, they can make you dizzy, they can make you nauseous, they are particularly good at keeping insects away (and some people, which is not necessarily a bad thing). Not much to do in terms of preparation. You could go outside and find an old rotten twig about the size of a cigar and practice holding it you mouth. For the uninitiated, the sensation is about the same. I am sure you will have ample opportunity to buy cigars Monday night. -- Doug > > On Nov 7, 2006, at 10:27 AM, Douglas Eadline wrote: > >> This is heart of the cigar world. Walt Ligon and I have been >> training six >> years for this. >> > > Well Doug, for us relative cigar novices, what advice would you offer > as to preparation for this event? I assume that there will be > opportunity for procurement of cigars for the event? > > I am already aware of the need for wearing sacrificial clothing ;-) > > -b > > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > From bill at Princeton.EDU Tue Nov 7 12:51:07 2006 From: bill at Princeton.EDU (Bill Wichser) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] LECCIBG Announcement (SC06) In-Reply-To: <1162929712.5766.5.camel@athlon64.site> References: <56148.192.168.1.1.1162913222.squirrel@mail.eadline.org> <771A122E-AC75-49C2-876F-F18827CB034E@ee.duke.edu> <1162929712.5766.5.camel@athlon64.site> Message-ID: <4550F1BB.6040105@princeton.edu> John, It's still tough to beat the prices here in Philly at Holt's http://www.holts.com Bill John Leidel wrote: > I would certainly suggest making a stop at the Thompsons Cigar factory > store on the way from the airport. > > http://maps.google.com/maps?f=q&hl=en&q=thompsons+cigars+near+Tampa, > +FL&ie=UTF8&z=13&ll=27.989854,-82.479515&spn=0.104895,0.161018&om=1&iwloc=A > > :-) > > cheers > john > > On Tue, 2006-11-07 at 14:23 -0500, Bill Rankin wrote: > >>On Nov 7, 2006, at 10:27 AM, Douglas Eadline wrote: >> >> >>>This is heart of the cigar world. Walt Ligon and I have been >>>training six >>>years for this. >>> >> >>Well Doug, for us relative cigar novices, what advice would you offer >>as to preparation for this event? I assume that there will be >>opportunity for procurement of cigars for the event? >> >>I am already aware of the need for wearing sacrificial clothing ;-) >> >>-b >> >>_______________________________________________ >>Beowulf mailing list, Beowulf@beowulf.org >>To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf > > > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf From bill at cse.ucdavis.edu Tue Nov 7 17:01:17 2006 From: bill at cse.ucdavis.edu (Bill Broadley) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <454DED19.3060300@scalableinformatics.com> References: <454DED19.3060300@scalableinformatics.com> Message-ID: <45512C5D.8090501@cse.ucdavis.edu> Joe Landman wrote: > We are working on a machine with 2 Opteron 2218s. For laughs, I ran > streams on it. Here are 1,2,3,4 processor data. I have a opteron 2220 handy with DDR2-667. At least I'm pretty sure it's DDR2-667. I multiplied the default N by a factor of 10 to get timings similar to yours. > 1 processor: > Copy: 5713.9944 0.0591 0.0560 0.0600 > Scale: 5713.9822 0.0587 0.0560 0.0600 > Add: 5454.2389 0.0911 0.0880 0.0920 > Triad: 5454.1576 0.0916 0.0880 0.0920 Function Rate (MB/s) Avg time Min time Max time Copy: 6615.1324 0.0484 0.0484 0.0484 Scale: 6665.6930 0.0480 0.0480 0.0481 Add: 6147.8462 0.0781 0.0781 0.0781 Triad: 6175.3782 0.0777 0.0777 0.0778 > 2 processors: > > Function Rate (MB/s) Avg time Min time Max time > Copy: 5713.8788 0.0591 0.0560 0.0640 > Scale: 5713.9822 0.0596 0.0560 0.0640 > Add: 5713.9538 0.0858 0.0840 0.0880 > Triad: 5713.9538 0.0871 0.0840 0.0880 Obviously using 2 sockets (not 2 cores on 1 socket): Function Rate (MB/s) Avg time Min time Max time Copy: 11836.5091 0.0270 0.0270 0.0270 Scale: 11685.2307 0.0274 0.0274 0.0274 Add: 11923.6815 0.0403 0.0403 0.0403 Triad: 11972.4659 0.0401 0.0401 0.0401 > 4 processors: > > Copy: 11428.1834 0.0293 0.0280 0.0320 > Scale: 11428.1347 0.0298 0.0280 0.0320 > Add: 10908.5812 0.0449 0.0440 0.0480 > Triad: 10908.6107 0.0440 0.0440 0.0440 Function Rate (MB/s) Avg time Min time Max time Copy: 12828.7019 0.0250 0.0249 0.0251 Scale: 12942.8860 0.0248 0.0247 0.0249 Add: 12910.1852 0.0374 0.0372 0.0377 Triad: 13030.0040 0.0369 0.0368 0.0372 > I did get one outlier at 4 CPUs. Could be a timing glitch. > > 4 processors: > Copy: 13332.7765 0.0302 0.0240 0.0360 > Scale: 13332.7765 0.0262 0.0240 0.0320 > Add: 13332.6220 0.0400 0.0360 0.0480 > Triad: 14999.0756 0.0391 0.0320 0.0480 Your timer does look dramatically less accurate than mine. The granularity of yours looks to be around 0.0040. Where mine is 0.0001 or so. My runs were pretty consistent. Notice that many of my min, max, and averages are the same. From greg.lindahl at qlogic.com Tue Nov 7 18:34:04 2006 From: greg.lindahl at qlogic.com (Greg Lindahl) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <200611071334.25659.csamuel@vpac.org> References: <454DED19.3060300@scalableinformatics.com> <454E75F1.4010202@scalableinformatics.com> <20061106182737.GH1892@greglaptop.rchland.ibm.com> <200611071334.25659.csamuel@vpac.org> Message-ID: <20061108023404.GA3339@greglaptop.internal.keyresearch.com> On Tue, Nov 07, 2006 at 01:34:25PM +1100, Chris Samuel wrote: > Have they moved to FlexLM yet ? We had PathScale on eval for a while but > found that their bespoke license daemon hung onto a users license for a long > while after they'd finished compiling whereas the Intel compilers just > worked.. This is a policy issue, and has nothing to do with moving to FlexLM. And it's discussed in the documentation. The policy is that once a user has a license, they have it for 15 minutes. If you need more licenses, buy more licenses. If you need more eval licenses, just ask. Intel probably has a more lenient policy for their compilers, but do keep in mind that their main business is selling chips, not selling software. So economics apply quite differently to them. -- greg From hahn at physics.mcmaster.ca Tue Nov 7 19:43:47 2006 From: hahn at physics.mcmaster.ca (Mark Hahn) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <20061108023404.GA3339@greglaptop.internal.keyresearch.com> References: <454DED19.3060300@scalableinformatics.com> <454E75F1.4010202@scalableinformatics.com> <20061106182737.GH1892@greglaptop.rchland.ibm.com> <200611071334.25659.csamuel@vpac.org> <20061108023404.GA3339@greglaptop.internal.keyresearch.com> Message-ID: >> Have they moved to FlexLM yet ? We had PathScale on eval for a while but >> found that their bespoke license daemon hung onto a users license for a long >> while after they'd finished compiling whereas the Intel compilers just >> worked.. > > This is a policy issue, and has nothing to do with moving to > FlexLM. And it's discussed in the documentation. > > The policy is that once a user has a license, they have it for 15 > minutes. If you need more licenses, buy more licenses. If you need > more eval licenses, just ask. I agree with Pathscale on this. we evaluated Pathscale about 2 years ago, and were rather apalled at the license-hold time (which was substantially longer at the time.) the reason I think 15 minutes is reasonable is that the value of the compiler+support does not reduce as computers speed up, making the compiler run faster. I would WAG that most compilations (of a single source file) run in <100 ms - probably 50x faster than a few year ago. Pathscale's license-hold time was previously much longer - even it made some sense, since for a set of full-time sw developers, it was a reasonable and more flexible way of approximating a per-seat license. our environment is academic HPC, in which compilation is much more bursty, and for which 15 minutes is really quite appropriate if you goal is to pay something that approximates the marginal utility of the license. obviously, there are alternatives ranging down to zero cost. we have had 20 Pathscale licenses (for ~11 clusters, 1400 total users, 7k opteron cores) for a bit over a year and are satisfied. regards, mark hahn. From csamuel at vpac.org Tue Nov 7 23:32:00 2006 From: csamuel at vpac.org (Chris Samuel) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <20061108023404.GA3339@greglaptop.internal.keyresearch.com> References: <454DED19.3060300@scalableinformatics.com> <200611071334.25659.csamuel@vpac.org> <20061108023404.GA3339@greglaptop.internal.keyresearch.com> Message-ID: <200611081832.03966.csamuel@vpac.org> On Wednesday 08 November 2006 13:34, Greg Lindahl wrote: > The policy is that once a user has a license, they have it for 15 > minutes. If you need more licenses, buy more licenses. If you need > more eval licenses, just ask. Understood, but the economics were sadly just not there for us in this case as we already had the Intel licenses for our other cluster and it was trivial to point the Opteron cluster to the same server. Being a non-profit academic research support organisation with many hundreds of users we have to cut our cloth to fit. In a perfect world we would have licenses for every compiler around... > Intel probably has a more lenient policy for their compilers, but do > keep in mind that their main business is selling chips, not selling > software. So economics apply quite differently to them. Agreed. Sorry if it came across a bit combatative, it wasn't meant to.. cheers! Chris -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available Url : http://www.scyld.com/pipermail/beowulf/attachments/20061108/3eefd841/attachment.bin From csamuel at vpac.org Tue Nov 7 23:33:17 2006 From: csamuel at vpac.org (Chris Samuel) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: References: <454DED19.3060300@scalableinformatics.com> <20061108023404.GA3339@greglaptop.internal.keyresearch.com> Message-ID: <200611081833.17621.csamuel@vpac.org> On Wednesday 08 November 2006 14:43, Mark Hahn wrote: > I agree with Pathscale on this. ?we evaluated Pathscale about 2 years ago, > and were rather apalled at the license-hold time (which was substantially > longer at the time.) Aha, that was around the time that we looked at it last. -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia From todd at acpdata.com Tue Nov 7 08:29:20 2006 From: todd at acpdata.com (Todd Patton) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] LECCIBG Announcement (SC06) In-Reply-To: <56148.192.168.1.1.1162913222.squirrel@mail.eadline.org> References: <56148.192.168.1.1.1162913222.squirrel@mail.eadline.org> Message-ID: <1162916960.11487.9.camel@jupiter.acpdata> For that matter; should anyone need some information, help, or assistance while in Tampa. Please feel free to email me. Although I am unable to attend the conference, I live in Tampa.On Tue, 2006-11-07 at 10:27 -0500, Douglas Eadline wrote: > For those attending SC06 and who want to make there cloths smell bad. > > The SC06 LECCIBG will be at the King Corona Cigars Bar/Club located at > 1523 E 7th Ave Ybor City. The corner of N 15th and E 7th. We will start > about 9:30 PM or so after the opening Gayla. > > More here: > > http://www.clustermonkey.net//content/view/164/57/ > > This is heart of the cigar world. Walt Ligon and I have been training six > years for this. > For that matter; should anyone need some information, help, or assistance while in Tampa, (aside from bail money), please feel free to email me. Although I am unable to attend the conference, I live in Tampa and know the area quite well. -- Todd Patton -----BEGIN PGP PUBLIC KEY BLOCK----- Version: GnuPG v1.4.2 (GNU/Linux) mQGiBEVPhocRBADnmDhjjxnoVNY3fY6UtkUdQ8dZtMkjOHhZOriP0kliQMJ4e+6M VzZdJiIH2qAXKzKAvkt6uGFOqwI3haDtBqrgl4yKVZd8YRWRlfCw34Q+jJNlzpRa Uhbyq8tGSJmI/KHnd4odDwXCC2golBS8/LT1mghkcnio1Qu7g7m0DejaowCg5FM1 aYmKRGalsMReJeTrr2rNQIsEAMgnmYyP3/Dc2QQliCy7U+W2qZkENr3ZOQIhOk7M H+l+lmSjk0iUk3bvOFdblRsSWUxImo0i3stpZnFEnDQ8akiNR9F8kRnL5rftNTUA 418TFvfvwjvM8iMUka8jMyiuNezh58YD8iqqXMcp7oh5IO0648N9NR3SOVR+QVek IjwBA/99avAkiH4G5GUsEMXHYNTNi5J+vtlnjCDmUk9Ykfx/siUmvzDPZ3++SZym JzKkSGKdcb+DLp6bRhCPt+lWCSfZisL1yVFKTJrNQ5NYTwOgq43Ir6LPqKKqL+za Bm2ZfpXDHZKBBCS0/vcfGRJil/eeps19KvqPbSdYTANoPSVmGbQhVG9kZCBELiBQ YXR0b24gPFRvZGRAYWNwZGF0YS5jb20+iGYEExECACYFAkVPhocCGwMFCQBPGgAG CwkIBwMCBBUCCAMEFgIDAQIeAQIXgAAKCRASYxowRTGgvUlaAJ9jI5qjELlLHFW7 p7s+ZEYVlXBKEwCfXsEcwXqb3zb+1N0W7Pq8Kia1Hw+5Ag0ERU+GqhAIANlM3lwj jAdY+AEF4TS2WHfNAuerjQewb8t7s6ZPwJ53P8bcFcD4s6z+JDOxz9WsQV1XTweC p8htCS7UCBPnaA7vERqzln9sF6vgC0M7CL/Onr/DgiITS+ThJd5HY1Emd8YKDoEc NAXO04QgPARr2pQxCr47G4YeIfv25sVJGE+8o94Lj9bjRr3w961gO4nzCjIwhusR wmtwHamFKMLpVPFUCslePGVFmmLJf3z/22hpPKsvHV7RhsCQU+WhXh0b+e99d1M9 QY+6Td4MMjAPbXr9n9FE34GNEfP+2j2VScPCT8lVMiSru5zKt9dLnTkJrR1oIxSx 6JqK/BWQpogG0l8ABREH/3MZVL8uEKANv458kPQ5/DbkbL/ItttbiNN+M5+E2Id5 lPcub1d+rfYWRfpsxCmCYOrM7Vh7QMLLdTyBsq/Jq4TrmA/Ojlf6dF80vAB+HOOF j97hcFfC/eeXRm9AulvlAxRGUTmQ0Emq1RDSfJNwsFGMip/7kZ8KnVBp6PTM1Rop Uqhi30bsd1Odm2WFQxZloM3ZGdyqngi8sMlma35QhrRrr6DMqNFjuRhtDhAcb3Nr 0DgaRRAOiT2ybug3bBc9kCbrxRoZy/eLIx3uWln6zcCM9Qi2QbMsPg7F6pKZjtQU mqaaJSeJIVghMd+xgWWVu28NAZtt8sVVAGEWOMnI1nKITwQYEQIADwUCRU+GqgIb DAUJAE8aAAAKCRASYxowRTGgvVKJAJ982NV3i/KP3FDoWXCEn62rO81hKwCcCh+2 eZhAEyHGlvtxO1lMm4IiTno= =Mz/h -----END PGP PUBLIC KEY BLOCK----- From landman at scalableinformatics.com Wed Nov 8 10:40:20 2006 From: landman at scalableinformatics.com (Joe Landman) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] OT: HPC technical position open Message-ID: <45522494.2050808@scalableinformatics.com> Hi folks: Please forgive the OT post / spam. Short version: we have an on-site HPC position open. Email me offline or call if you want more info. If you will be at SC06, my cell is below, please call, and we can meet and talk. Joe -- Joseph Landman, Ph.D Founder and CEO Scalable Informatics LLC, email: landman@scalableinformatics.com web : http://www.scalableinformatics.com phone: +1 734 786 8423 fax : +1 734 786 8452 or +1 866 888 3112 cell : +1 734 612 4615 From greg.lindahl at qlogic.com Wed Nov 8 14:10:03 2006 From: greg.lindahl at qlogic.com (Greg Lindahl) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Thought that this might be of interest In-Reply-To: <200611081832.03966.csamuel@vpac.org> References: <454DED19.3060300@scalableinformatics.com> <200611071334.25659.csamuel@vpac.org> <20061108023404.GA3339@greglaptop.internal.keyresearch.com> <200611081832.03966.csamuel@vpac.org> Message-ID: <20061108221003.GB4871@greglaptop.internal.keyresearch.com> On Wed, Nov 08, 2006 at 06:32:00PM +1100, Chris Samuel wrote: > Understood, but the economics were sadly just not there for us in this case as > we already had the Intel licenses for our other cluster and it was trivial to > point the Opteron cluster to the same server. Yep, it's hard to compete with software you've already paid for :-) -- greg From csamuel at vpac.org Wed Nov 8 18:51:45 2006 From: csamuel at vpac.org (Chris Samuel) Date: Fri May 9 01:05:27 2008 Subject: NFS Performance (was Re: [Beowulf] GPFS on Linux (x86)) In-Reply-To: References: <32fc08be0609111914u708ce1aev4c86be4e7e6e4fe4@mail.gmail.com> <200609171321.45829.csamuel@vpac.org> Message-ID: <200611091351.45785.csamuel@vpac.org> On Tuesday 19 September 2006 19:55, Krugger wrote: > Could it be that you are getting timeouts because you don't have > enough nfsd daemons. This only might happen under heavy I/O load. And > when there are thousands of operations floading the server sometimes > it just has to drop some off the queue when the queue get full. I had the number of NFS daemons up to around 128 at one point. That only succeeded in getting us a load average of 90+ with them all sitting around in device waits. FC3 with XFS worked just fine with the standard number of daemons. cheers, Chris (catching up on old email) -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia From csamuel at vpac.org Wed Nov 8 18:57:26 2006 From: csamuel at vpac.org (Chris Samuel) Date: Fri May 9 01:05:27 2008 Subject: NFS Performance (was Re: [Beowulf] GPFS on Linux (x86)) In-Reply-To: <32fc08be0609150732x265158c9r9272d570a32655ac@mail.gmail.com> References: <32fc08be0609111914u708ce1aev4c86be4e7e6e4fe4@mail.gmail.com> <200609160005.53412.csamuel@vpac.org> <32fc08be0609150732x265158c9r9272d570a32655ac@mail.gmail.com> Message-ID: <200611091357.30928.csamuel@vpac.org> On Saturday 16 September 2006 00:32, Brent Franks wrote: > On 9/15/06, Chris Samuel wrote: > > > Wild guess - NFS servers are running RHEL and using ext3 ? > > Exactly. If you are using RHEL4 and are not at Update 4 already then apparently there is a RH induced bug in the default cfq io_scheduler and the work around is (according to Oracle at least [1]) to set the node to use the deadline scheduler at boot time as well. The better fix is apparently to update to RHEL4 U4. [1] - Their OCFS2 page says: # RHEL4 U2 and U3 users are advised to use the DEADLINE io scheduler instead # of the default CFQ due to Bugzilla# 184535. (Ed: [2]) [2] - https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=184535 cheers, Chris -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available Url : http://www.scyld.com/pipermail/beowulf/attachments/20061109/a99d0bb8/attachment.bin From csamuel at vpac.org Wed Nov 8 19:04:00 2006 From: csamuel at vpac.org (Chris Samuel) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Apologies for the spam/virus yesterday In-Reply-To: <20060209002848.GA16857@cse.ucdavis.edu> References: <20060209002848.GA16857@cse.ucdavis.edu> Message-ID: <200611091404.03489.csamuel@vpac.org> On Thursday 09 February 2006 11:28, Bill Broadley wrote: > I'd vote for banning ALL posts with attachments, HTML, vcard, .exe. > > Has there ever been a useful attachment sent to the beowulf list? PGP signatures.. -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available Url : http://www.scyld.com/pipermail/beowulf/attachments/20061109/1a0c75b9/attachment.bin From csamuel at vpac.org Wed Nov 8 19:13:16 2006 From: csamuel at vpac.org (Chris Samuel) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Blue-sky cluster security [was CLuster - Mpich - tstmachines - Heeelp !!!!!!!!] In-Reply-To: <44CAD0CA.8070003@gmail.com> References: <4c808ef0607101800y4a97363fw889b55daa7be8b0b@mail.gmail.com> <44CAD0CA.8070003@gmail.com> Message-ID: <200611091413.16806.csamuel@vpac.org> On Saturday 29 July 2006 13:06, Geoff Jacobs wrote: > Just brainstorming, what would the best method be? Poach some of the > globus stuff? A chroot scheme? Xen? The Globus folks have some work that I saw described at CCGrid in Singapore this year called Workspaces [1] which is basically deploying a Xen VM to provide a known OS and application stack for grid jobs. It has the added advantage that once the job is complete and results retrieved then the VM can be destroyed and it is out of the way. They have separate filesystems for the root, usr and other areas to make it possible to mix and match. The driver behind this is the age old question of grid computing that if I can submit a job and a meta-scheduler picks a random cluster, how can I know that it will have the right environment for my application ? [1] - http://workspace.globus.org/ Chris -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia From mwill at penguincomputing.com Thu Nov 9 08:37:51 2006 From: mwill at penguincomputing.com (Michael Will) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Apologies for the spam/virus yesterday Message-ID: <433093DF7AD7444DA65EFAFE3987879C24545C@jellyfish.highlyscyld.com> Completely usel.ess unless you plan to litigate and have to prove authorship -----Original Message----- From: Chris Samuel [mailto:csamuel@vpac.org] Sent: Wed Nov 08 19:20:28 2006 To: beowulf@beowulf.org Subject: Re: [Beowulf] Apologies for the spam/virus yesterday On Thursday 09 February 2006 11:28, Bill Broadley wrote: > I'd vote for banning ALL posts with attachments, HTML, vcard, .exe. > > Has there ever been a useful attachment sent to the beowulf list? PGP signatures.. -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.scyld.com/pipermail/beowulf/attachments/20061109/528e6b63/attachment.html From rgb at phy.duke.edu Thu Nov 9 10:12:53 2006 From: rgb at phy.duke.edu (Robert G. Brown) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Apologies for the spam/virus yesterday In-Reply-To: <433093DF7AD7444DA65EFAFE3987879C24545C@jellyfish.highlyscyld.com> References: <433093DF7AD7444DA65EFAFE3987879C24545C@jellyfish.highlyscyld.com> Message-ID: On Thu, 9 Nov 2006, Michael Will wrote: > Completely usel.ess unless you plan to litigate and have to prove authorship I generally concur -- besides, signatures don't have to be attachments and one COULD except them from a rule against attachments anyway. However, attachment replication on a list is very resource intensive -- you generate NxB bytes of data in mailboxes everywhere, most of which is never even looked at. It was my understanding that the proper netiquette was to put the object you wish to distribute up on a webserver somewhere and then just post the URL. That makes it a client-pull action only by people who are interested, not a server-push action, which conserves resources all around. I very much doubt that anybody on this list with an attachment they WANT to distribute will not have a webserver accessible to facilitate the distribution. I also can't stand html formatted email messages in general (not messages with links, messages that can only be "read" in a browser) and my spam filter doesn't like them either, just in case anyone on list ever sends me such a hyperformatted message and wonders why I don't respond...;-) rgb > > -----Original Message----- > From: Chris Samuel [mailto:csamuel@vpac.org] > Sent: Wed Nov 08 19:20:28 2006 > To: beowulf@beowulf.org > Subject: Re: [Beowulf] Apologies for the spam/virus yesterday > > On Thursday 09 February 2006 11:28, Bill Broadley wrote: > >> I'd vote for banning ALL posts with attachments, HTML, vcard, .exe. >> >> Has there ever been a useful attachment sent to the beowulf list? > > PGP signatures.. > > -- Robert G. Brown http://www.phy.duke.edu/~rgb/ Duke University Dept. of Physics, Box 90305 Durham, N.C. 27708-0305 Phone: 1-919-660-2567 Fax: 919-660-2525 email:rgb@phy.duke.edu From landman at scalableinformatics.com Thu Nov 9 14:09:47 2006 From: landman at scalableinformatics.com (Joe Landman) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Slightly OT: storage performance Message-ID: <4553A72B.1040203@scalableinformatics.com> Hi folks: I am running some IOzone (and bonnie++, and spew, and we wrote our own RAFAP*), and I want to get a feel for what people consider "good" performance. I know IOzone isn't really representative of workloads, and I personally abhor "benchmarks" which don't make at least an effort to reflect real workloads (which is an issue in and of itself, as there aren't any "standard" workloads that I am aware of for servers, everyones will be different ...). My question is this: apart from using huge file sizes to see raw disk performance, what do you considered good performance on the various tests, either in the huge file size regime, or in the cache interaction regime? Basically which tests are most meaningful to your workloads? Are the raw disk data really the most useful datum? Are they corner cases that you are simply interested in? Is the most important test case reading and changing one byte at random in a 1TB file, several hundred million times? Or is it large block sequential IO? Disclosure: working on a white paper for something we are working on (bug me at SC), and I eschew using completely meaningless numbers. If there is some sort of cutoff that people have between what they consider "eh" and "good", I would like to hear it. You can email me offline if you want, and I will summarize later on. Thanks. Joe * RAFAP == Read As Fast As Possible, a really simple C code that tries to hammer on the IO. -- Joseph Landman, Ph.D Founder and CEO Scalable Informatics LLC, email: landman@scalableinformatics.com web : http://www.scalableinformatics.com phone: +1 734 786 8423 fax : +1 734 786 8452 or +1 866 888 3112 cell : +1 734 612 4615 From csamuel at vpac.org Thu Nov 9 19:43:23 2006 From: csamuel at vpac.org (Chris Samuel) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Apologies for the spam/virus yesterday In-Reply-To: <433093DF7AD7444DA65EFAFE3987879C24545C@jellyfish.highlyscyld.com> References: <433093DF7AD7444DA65EFAFE3987879C24545C@jellyfish.highlyscyld.com> Message-ID: <200611101443.27995.csamuel@vpac.org> On Friday 10 November 2006 03:37, Michael Will wrote: > Completely usel.ess unless you plan to litigate and have to prove > authorship Certainly for me being able to demonstrate authorship and/or modification of a message or file is useful in its own right, I don't understand why utility should be considered conditional upon legal action. E.g. LISTSERV used to (may still do) convert tabs to spaces in messages passing through it, so if someone creates a file or patch where tabs are significant and OpenPGP/MIME signs it and sends it to a LISTSERV list then the signature would fail and the recipients will know the something is wrong with it. Chris -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available Url : http://www.scyld.com/pipermail/beowulf/attachments/20061110/3cd13e02/attachment.bin From csamuel at vpac.org Thu Nov 9 20:11:47 2006 From: csamuel at vpac.org (Chris Samuel) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Apologies for the spam/virus yesterday In-Reply-To: References: <433093DF7AD7444DA65EFAFE3987879C24545C@jellyfish.highlyscyld.com> Message-ID: <200611101511.50763.csamuel@vpac.org> On Friday 10 November 2006 05:12, Robert G. Brown wrote: > besides, signatures don't have to be attachments That's the old style ASCII armour technique, effectively supersceded in 1996 by RFC 2015 (based on the previous PGP/MIME work). -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available Url : http://www.scyld.com/pipermail/beowulf/attachments/20061110/6e213738/attachment.bin From csamuel at vpac.org Thu Nov 9 20:23:32 2006 From: csamuel at vpac.org (Chris Samuel) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Slightly OT: storage performance In-Reply-To: <4553A72B.1040203@scalableinformatics.com> References: <4553A72B.1040203@scalableinformatics.com> Message-ID: <200611101523.35651.csamuel@vpac.org> On Friday 10 November 2006 09:09, Joe Landman wrote: > ? ?My question is this: ?apart from using huge file sizes to see raw > disk performance, what do you considered good performance on the various > tests, either in the huge file size regime, or in the cache interaction > regime? ?Basically which tests are most meaningful to your workloads? > Are the raw disk data really the most useful datum? Because of our diversity of users (500+ from 8 separate universities covering a myriad of fields) we can't really been able to pick anything better than Bonnie++ so far. The raw data does seem to give a generally reasonable indication of how well a system can do I/O, for us at least. cheers, Chris -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available Url : http://www.scyld.com/pipermail/beowulf/attachments/20061110/c8523c2c/attachment.bin From csamuel at vpac.org Fri Nov 10 14:04:47 2006 From: csamuel at vpac.org (Chris Samuel) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] More cores/More processors/More nodes? In-Reply-To: <36384.192.168.1.1.1159886250.squirrel@mail.eadline.org> References: <1159466774.3038.14.camel@ceiriog.eclipse.co.uk> <20060930105758.GC5913@greglaptop> <36384.192.168.1.1.1159886250.squirrel@mail.eadline.org> Message-ID: <200611110904.52300.csamuel@vpac.org> On Wednesday 04 October 2006 00:37, Douglas Eadline wrote: > This is a non-obvious result many find hard to believe. > That is, MPI on the same node maybe faster than some shared/threaded > mode. (of course it all depends on the application etc.) We believe we have seen that with LS-Dyna comparing SMP versus MPP (MPI) variants on a single node. cheers, Chris -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available Url : http://www.scyld.com/pipermail/beowulf/attachments/20061111/4e97988a/attachment.bin From landman at scalableinformatics.com Fri Nov 10 14:52:48 2006 From: landman at scalableinformatics.com (Joe Landman) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] More cores/More processors/More nodes? In-Reply-To: <200611110904.52300.csamuel@vpac.org> References: <1159466774.3038.14.camel@ceiriog.eclipse.co.uk> <20060930105758.GC5913@greglaptop> <36384.192.168.1.1.1159886250.squirrel@mail.eadline.org> <200611110904.52300.csamuel@vpac.org> Message-ID: <455502C0.6060007@scalableinformatics.com> I must have missed the original message. Chris Samuel wrote: > On Wednesday 04 October 2006 00:37, Douglas Eadline wrote: > >> This is a non-obvious result many find hard to believe. >> That is, MPI on the same node maybe faster than some shared/threaded >> mode. (of course it all depends on the application etc.) > > We believe we have seen that with LS-Dyna comparing SMP versus MPP (MPI) > variants on a single node. Actually (my apologies to Doug for not seeing this before and commenting), the Cray folks have noted this for a number of years, specifically when they moved MPI code from T3D/T3E back onto Vector boxes, it behaved better. The rationale for this is that MPI does in fact impose a very stringent restriction on sharing in that you must explicitly share data. This *forces* you into spatial localization, and to a degree, temporal localization, as you have to get the most out of your data. OpenMP (and please don't beat me up on this: I like the OpenMP syntax/usage model better than MPI) doesn't force you to explicitly share items. The compiler does it for you. If you forget, well, thats ok, it will default to doing the right thing. This also means that you may not work as hard to spatially localize your computation as you would in MPI. MPI requires a more rigid discipline, and if you have ever debugged an MPI program, it also requires lots of antacid. In the age of gui debuggers, I find myself using print statements, and barriers like MPI_Barrier... (yeah I know about Etnus, but I don't have enough need to justify the purchase). I don't have enough experience with PGAS languages to see whether or not they provide the ease of use of OpenMP with the rigid sharing and localization of MPI. I still think Chapel looks neat, would love to play with a compiler for it. I think UPC might even run on some clusters, last I looked you had to compile it remotely (some sort of web page thingy). Joe > > cheers, > Chris > > > ------------------------------------------------------------------------ > > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf -- Joseph Landman, Ph.D Founder and CEO Scalable Informatics LLC, email: landman@scalableinformatics.com web : http://www.scalableinformatics.com phone: +1 734 786 8423 fax : +1 734 786 8452 cell : +1 734 612 4615 From johannesrs at gmail.com Wed Nov 8 19:45:31 2006 From: johannesrs at gmail.com (Jones de Andrade) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Core 2 Duo Bugs: Any importance in the beowulf field? Message-ID: <54e4355e0611081945o7219bfbbg4d2d20062b3920e@mail.gmail.com> Hi all. Wel, let me go straigh to the point: I'm a begginer, more or less enthusiastic about the beowulf world. I've been keeping track of this list for a good time. I would like to aks just one single (and probably stupid) question: A few months ago, Intel released a small list of bugs of the first version of the Core 2 chips family. Much discussion has gone around the web about whether or not these bugs (or some, or any of them) would affect any class of users. I haven't seen such a discussion here. I would like to know if there aren't any concerns on using this chips in the kind ofapplication beowulfs are usually used? Specially for scientific applications? And what's the experience anyone got with those chips? Good, bad... Thanks a lot in advance, and sorry for wasting your tie, Sincerally yours, Jones -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.scyld.com/pipermail/beowulf/attachments/20061109/fb86bf71/attachment.html From m.dierks at skynet.be Fri Nov 10 16:27:13 2006 From: m.dierks at skynet.be (Michel Dierks) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] MPI synchronization problem Message-ID: <455518E1.2060006@skynet.be> Hello, I'm developing an application to calculate the Mandelbrot fractal. My problem is: from the master I will send to each free node for computation a message containing the values of one screen line. After a time the nodes will send back the results. I have see that I can use MPI_Isend and MPI_Irecv for non-blocking communication. This give me no problem if we are talking about of one send and one receive. But how can I handle the sending from each node to the master without data corruption. Must I implement a send and a receive buffer for each node (16 nodes means 16 buffer in and 16 out) ? Can someone help me? Please From gdjacobs at gmail.com Fri Nov 10 20:42:54 2006 From: gdjacobs at gmail.com (Geoff Jacobs) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] MPI synchronization problem In-Reply-To: <455518E1.2060006@skynet.be> References: <455518E1.2060006@skynet.be> Message-ID: <455554CE.9080500@gmail.com> Michel Dierks wrote: > Hello, > > I'm developing an application to calculate the Mandelbrot fractal. Hmm... I never did this when I was learning the ropes. There is a pretty example included with the mpich source code to do Mandelbrot fractals if you need something to crib from. > My problem is: > > from the master I will send to each free node for computation a message > containing the values of one screen line. I would think you can get away with transmitting just the boundaries of the section of complex plane which you will be plotting, as well as the iterative limit, color interpretation, etc. Slaves can determine what portion they must calculate from their ranks. > After a time the nodes will send back the results. > > I have see that I can use MPI_Isend and MPI_Irecv for non-blocking > communication. This give me no problem if we are talking about of one > send and one receive. But how can I handle the sending from each node to > the master without data corruption. > Must I implement a send and a receive buffer for each node (16 nodes > means 16 buffer in and 16 out) ? The software I've implemented has tended to be rather dynamic, so it seemed easier to use discrete sends and receives. In your case, you will be calculating a vector of fixed size on each slave (one row for your resultant image per). It would be logical to use a collective communicator like MPI_Gather to automatically populate the array representing your raster image. http://csit1cwe.fsu.edu/extra_link/pe/d3a64mst07.html#HDRIGATH > Can someone help me? Please > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > -- Geoffrey D. Jacobs Go to the Chinese Restaurant, Order the Special From deadline at clustermonkey.net Sat Nov 11 16:15:31 2006 From: deadline at clustermonkey.net (Douglas Eadline) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Core 2 Duo Bugs: Any importance in the beowulf field? In-Reply-To: <54e4355e0611081945o7219bfbbg4d2d20062b3920e@mail.gmail.com> References: <54e4355e0611081945o7219bfbbg4d2d20062b3920e@mail.gmail.com> Message-ID: <50731.192.168.1.1.1163290531.squirrel@mail.eadline.org> Generally these issues are handled by compilers, BIOS, microcode updates, and steppings of the processor. All new processors have a list of bugs. The AI 39 seems to be of some concern. Check this document, http://www.intel.com/design/processor/specupdt/313279.htm Again, don't be alarmed this is standard for all processors. -- Doug > Hi all. > > Wel, let me go straigh to the point: I'm a begginer, more or less > enthusiastic about the beowulf world. I've been keeping track of this list > for a good time. > > I would like to aks just one single (and probably stupid) question: A few > months ago, Intel released a small list of bugs of the first version of > the > Core 2 chips family. Much discussion has gone around the web about whether > or not these bugs (or some, or any of them) would affect any class of > users. > > I haven't seen such a discussion here. I would like to know if there > aren't > any concerns on using this chips in the kind ofapplication beowulfs are > usually used? Specially for scientific applications? > > And what's the experience anyone got with those chips? Good, bad... > > Thanks a lot in advance, and sorry for wasting your tie, > > Sincerally yours, > > Jones > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > -- Doug From deadline at clustermonkey.net Sat Nov 11 17:20:37 2006 From: deadline at clustermonkey.net (Douglas Eadline) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] Late News: SC06 Panel In-Reply-To: <60043.192.168.1.1.1162404914.squirrel@mail.eadline.org> References: <200611011815.22799.csamuel@vpac.org> <60043.192.168.1.1.1162404914.squirrel@mail.eadline.org> Message-ID: <53967.192.168.1.1.1163294437.squirrel@mail.eadline.org> The geographically challenged San Francisco Bay Area Beowulf Users Group (BayBUG) would like to invite everyone to attend an exclusive SC06 panel discussion before the BeoBash (Wednesday November 14th) entitled The Commercialization of Open Source - for better or worse? Panelists: * Donald Becker, CTO of Penguin Computing, Beowulf cluster co-inventor * Thomas Sterling, Faculty Associate, Center for Advanced Computing Research, California Institute of Technology (Caltech) Beowulf cluster co-inventor * Mike Fitzmaurice, HPC Technologist at GTSI Corp. The panel will run from 4:00 - 5:30 p.m and be held at the Rock-N-Sports Bar and Grille. The BeoBash starts afterward at 6:00PM Details Here: http://www.clustermonkey.net//content/view/170/2/ -- Doug From deadline at clustermonkey.net Sat Nov 11 17:22:44 2006 From: deadline at clustermonkey.net (Douglas Eadline) Date: Fri May 9 01:05:27 2008 Subject: [Beowulf] SC'06 Beowulf Bash ? In-Reply-To: <60043.192.168.1.1.1162404914.squirrel@mail.eadline.org> References: <200611011815.22799.csamuel@vpac.org> <60043.192.168.1.1.1162404914.squirrel@mail.eadline.org> Message-ID: <53974.192.168.1.1.1163294564.squirrel@mail.eadline.org> Jim Cownie mentioned that there is a "Trolley" and a "Street Car" The Streetcar goes to Ybor City from the convention center (not the Trolley): http://www.tecolinestreetcar.org/about/maps/index.htm This map seems to confirm the Streetcar route. Any one who has been/lives in Tampa have any comment? Thanks -- Doug > > Beobash details: > > http://www.clustermonkey.net//content/view/166/2/ > > -- > Doug > > >> Hi folks, >> >> I'm off to my first SC'06 this year to help man the VPAC booth with a >> couple >> of friends and was wondering if there were any details about the Beowulf >> bash >> yet ? >> >> cheers, >> Chris >> -- >> Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager >> Victorian Partnership for Advanced Computing http://www.vpac.org/ >> Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia >> >> _______________________________________________ >> Beowulf mailing list, Beowulf@beowulf.org >> To change your subscription (digest mode or unsubscribe) visit >> http://www.beowulf.org/mailman/listinfo/beowulf >> > > > -- > Doug > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > -- Doug From csamuel at vpac.org Sat Nov 11 20:17:46 2006 From: csamuel at vpac.org (Chris Samuel) Date: Fri May 9 01:05:28 2008 Subject: [Beowulf] SC'06 Beowulf Bash ? In-Reply-To: <53974.192.168.1.1.1163294564.squirrel@mail.eadline.org> References: <200611011815.22799.csamuel@vpac.org> <60043.192.168.1.1.1162404914.squirrel@mail.eadline.org> <53974.192.168.1.1.1163294564.squirrel@mail.eadline.org> Message-ID: <200611121517.50907.csamuel@vpac.org> On Sunday 12 November 2006 12:22, Douglas Eadline wrote: > This map seems to confirm the Streetcar route. Any > one who has been/lives in Tampa have any comment? Can't comment on the streetcar, but the "trolley" bus (which has no actual trolley) is 50c for a ride and if you're going to be using it a lot you can get a 31 day pass for $5 from the company. To do that you'll need to take the 96 (not the 98) and ask the driver to show you where.. The 96 runs around the CBD every 15 minutes from around 6am to 6pm, and then every 30 minutes from 6pm to 9pm. The 98 goes out to Hyde Park and doesn't start running until 11:30am from memory.. cheers! Chris, wondering what the large raptors are that circle over Tampa CBD -- Christopher Samuel - (03)9925 4751 - VPAC Deputy Systems Manager Victorian Partnership for Advanced Computing http://www.vpac.org/ Bldg 91, 110 Victoria Street, Carlton South, VIC 3053, Australia -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available Url : http://www.scyld.com/pipermail/beowulf/attachments/20061112/2df462f4/attachment.bin From buccaneer at rocketmail.com Sun Nov 12 09:18:01 2006 From: buccaneer at rocketmail.com (Buccaneer for Hire.) Date: Fri May 9 01:05:28 2008 Subject: [Beowulf] Slightly OT: storage performance Message-ID: <20061112171801.31794.qmail@web30606.mail.mud.yahoo.com> ----- Original Message ---- From: Joe Landman [snip] > My question is this: apart from using huge file sizes to see raw > disk performance, what do you considered good performance on the various > tests, either in the huge file size regime, or in the cache interaction > regime? Basically which tests are most meaningful to your workloads? > Are the raw disk data really the most useful datum? Are they corner > cases that you are simply interested in? Is the most important test > case reading and changing one byte at random in a 1TB file, several > hundred million times? Or is it large block sequential IO? The good thing about using IOZone and Bonnie++ is that there is a large body of numbers that you can compare your equipment to. We also have a program written in-house that we have testing and verified over and over again. Besides the normal sequential read/write and single/multithread randoms, it can also mimic how seismic data is process in our real world. The minimum file size for testing should be at least twice the maximum estimated cache in the system. So, we use the amount of memory on the head node and all the cache on the EMC and double it. Then you start test and tune the fil