[Beowulf] Network considerations for new generation cheap beowulfcluster

Douglas Eadline deadline at eadline.org
Thu May 24 07:43:50 PDT 2007


> At 10:52 AM 5/23/2007, Peter St. John wrote:
>>But oh and Jim if you recall any papers about this I could read that
>>would be "Jim" Dandy.
>
> I was working off memory, and the iPSC/1 and iPSC/2 manuals I have in
> my office as a historical artifact.
>
> I seem to recall that if you google hypercube and intel, you'll turn
> up some of the papers that were written early on.  The guys who
> started with the hypercube interconnect were at CalTech, as I recall,
> and spun off to form a supercomputer company embodying that, which
> Intel also adopted.

nCUBE was a commercial hypercube machine. The first version
used software routing, so if a CPU crashed, so did the whole
machine. Version 2 had HW routing and after a hefty investment
by Larry Ellison. It was destined to become a parallel database machine
only it took a long time to get the Oracle working. As a result
nCUBE sold into the HPC market against TMC for a while. I used
to have an 8 processor board that plugged into a Sun 386i.
The last I read they were making video on-demand servers.

--
Doug


>
>>Peter
>>
>>
>>On 5/23/07, Jim Lux
>><<mailto:James.P.Lux at jpl.nasa.gov>James.P.Lux at jpl.nasa.gov> wrote:
>>At 09:19 AM 5/22/2007, Peter St. John wrote:
>>>A hypercube ( http://en.wikipedia.org/wiki/Hypercube) also gets you
>>>exponential space; the max hops is the dimension (3 for a
>>>3-dimensional cube) and the number of nodes is exp(base 2) of the
>>>dimension (8 vertices on a cube). To do a tesseract (4-cube), which
>>>looks like two cubes nested, you'd need 4 ports per node, 16 nodes,
>>>32 cables, max hop 4. I've poked around and don't see a great 4
>>>ports per node solution; I like the suggestion of putting a router
>>>on a motherboard.
>>
>>Mind you, this is what Intel started with on their iPSC/1 and iPSC/2
>>computers.  The early ones had multiple NICs in the nodes, then,
>>later, they had a 8 port (I think) router in each node.
>>
>>It's not clear that this saves anything over a simpler architecture
>>(e.g. external switch with lots of ports in a crossbar) unless you
>>can do circuit switched routing (so you don't have a one packet
>>delay in the switch) AND your algorithm can take advantage of it. I
>>spent quite some time in the late 80s trying to figure out clever
>>ways to take advantage of a hypercube topology for a modeling
>>application..  I'm sure there are algorithms which are a natural
>>fit, but the ones I was using weren't.
>>
>>
>>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
>>
>
> 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
>
> !DSPAM:46549197244853022792082!
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit
> http://www.beowulf.org/mailman/listinfo/beowulf
>
>
> !DSPAM:46549197244853022792082!
>


-- 
Doug



More information about the Beowulf mailing list