[Beowulf] Gentoo in the HPC environment

Joe Landman landman at scalableinformatics.com
Thu Jun 26 07:45:54 PDT 2014


On 06/26/2014 10:36 AM, Prentice Bisbal wrote:

>
> I disagree with needed separate images in a VM/Docker/etc. environment,
> maybe if your doing system-level research, but most HPC users work
> exclusively in the user-space,and just want to run their
> MATLAB/NAMD/LAMMPS, whatever job. In this case, just installing
> user-space applications and libraries in a different path from the
> distro-supplied versions is adequate. Modifying PATH and a few other

Not always ... I've seen library level collisions, where MKL runtime is 
supplied per node/OS, and the application assumes different versions. 
Similar for OpenMPI, HPMPI, IB, ...

This is where Docker shines.  Its a container, not a full VM.  Install 
what you need, and don't worry about stuff you don't.

Or run your machine as an entire container (e.g. diskless/ramdisk root), 
and solve the problem the same way.

> environment variables that most users can handle is all you need
> modules, lmod, softenv and other utilities make that even easier for users.

Sometimes ... Not all the time.  Limited sets of apps will play well 
with this.  Just start playing with things that assume that the first 
thing they find in the path named 'java' is the "right" one.  Or the 
first library in the tree managed by ld.so ...

The word 'nightmare' does not even begin to describe the debugging 
process for this.  Usually under the gun, with users needing the system 
operational for their needs in time scales measured in minutes at most.

Been there, done that, which is why things that completely solve the 
issue (e.g. shift the problem back to the user and their code) are a 
good thing.

> If you're doing lower-level research involving kernel modules or kernel
> tuning, yes, you will need VMs or something. But this is usually
> 'research', not 'production' HPC.
>
>>
>> I ended up doing very crazy root-stealing, chroot-establishing things
>> to get my science done in my PhD.  If you prevent intelligent people
>> from doing their work, they are going to be your worst nightmare.
>> Don't kid yourselves if you think you are doing anyone favors by
>> providing super-static OS environments like RHEL for your users.  You
>> are just being lazy (and not the good kind of programmer lazy).
>
> This is so true. If you are a roadblock to your users, they will find a
> way around you.

I was trying not to say that, as it alarms the people whom are steadfast 
in their belief in the "one true way" to deliver these services, but ... 
there is significant truth to this.

Put another way, then entire concept of beowulfery grew up around people 
dissatisfied with the diktat from on high on how to purchase/build/run 
their computing infrastructure.  They moved around the roadblocks.  Its 
an interesting object lesson to see what has happened to those roadblocks.

History may not necessarily precisely repeat itself, but it sure does 
rhyme a bit.


-- 
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics, Inc.
email: landman at scalableinformatics.com
web  : http://scalableinformatics.com
twtr : @scalableinfo
phone: +1 734 786 8423 x121
cell : +1 734 612 4615


More information about the Beowulf mailing list