[Beowulf] why we need cheap, open learning clusters
landman at scalableinformatics.com
Mon May 20 04:29:36 PDT 2013
On 05/20/2013 02:48 AM, Christopher Samuel wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> On 13/05/13 03:55, Lux, Jim (337C) wrote:
>> It starts to look like significant capital equipment, with only
>> authorized service, compliance with corporate/institutional IT
>> security rules: Do you have all your patches up to date?
> I guess that depends on whether your system is exposed to the outside
> world I'd say.
>> Are you running the institutional virus checker?.
> I once helped out a group of people in the mid 2000's who ran into
> just that issue - their organisations ITSEC SOPs mandated they run a
> virus checker once a day on their HPC cluster.
A long ... long time ago, on a blog far, far away
(http://scalability.org/?p=628) , I groused about what I called "IT
clusters". These were clusters designed and built by IT folks, not HPC
folks. Run the way IT folks would run it. Not the way HPC folks would
run it. Makes the IT crowd happy. End users? Not so much ...
> Fortunately the blessed virus checker had a Linux version, but
> unfortunately their fileserver was AIX. So they ended up running it
> on the management node over NFS.
> It wasn't a problem for a while until they had built up a reasonably
> amount of data (did I mention this was for bioinfomatics amongst other
> things?) and it started taking more than 24 hours to run it once.
> You can see where this is going can't you?
> Anyway, after a couple of management node meltdowns they managed to
> get their IT security folks to accept that perhaps they could get away
> with compiling ClamAV for AIX and running it on the file server itself
> where it could do minimal damage...
I hate to say this is the algorithm, but I've seen it so ... many ...
1) set something up (don't bring in them thar expensive experts, do it
2) run it until you run into problems
3) call in one of them thar expensive experts, but only for a short
time, to analyze what we are doing, and make suggestions
4) implement only the least cost (if any implemented at all)
5) Goto 2.
This won't change until there is a change in thinking, which either
means an enlightened group/manager/VP, or the barrier to doing the right
thing leaves/gets promoted into another position outside of the loop/...
Momentum is very easy in an IT group. Virus checkers? Sure, run them
everywhere. Even where they aren't needed. Because its policy, thats
why. Few of them take the step back, ask "why do we need virus
checkers", and start to work on solving the real problem (using
vulnerable and poorly designed systems).
Solving problems is great. Finding, thinking about, and solving the
right problems is even better.
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics, Inc.
email: landman at scalableinformatics.com
web : http://scalableinformatics.com
phone: +1 734 786 8423 x121
fax : +1 866 888 3112
cell : +1 734 612 4615
More information about the Beowulf