Clusters (was CLUG meeting 23 May 2002)

[Drake Diedrich]

On Thu, May 23, 2002 at 11:12:40AM +1000, Doug ABERDEEN wrote:
> 
> A grid is not quite the same as a cluster. The processing nodes of a grid
> can be distributed across a wide geographic area, whereas clusters are

> 
> When you think grid, think global cluster, with global organisational
> problems to go along.

   This is nothing new.  In 1990, before the term "Grid" existed, I was
using workstations and headless nodes distributed around campus (psu.edu),
under common authentication and using shared filespace.  Kerberos and AFS
were used, and were as secure as anything in use today.  DQS was the
resource allocation software, and was only used locally (though local
spanned around 100 nodes in a dozen departments).  The term used locally and
in the literature at that time was cluster.  "sites" were another term, used
for the campus-wide authentication systems that were then shared globally to
the AFS system.  The filespace and authentication ssytems were ultimately
worldwide, though in practice that meant about 20 campuses and 2-3
multinationals on the eastern seaboard of the U.S.  As the web started up,
there were even people posting file:///afs/x.y.edu/... URLs, as everyone who
mattered had the common global /afs directory anyway, no need for that
noisy, inefficient, uncacheable HTTP protocol.  :)

Whatever "Grid" means, it doesn't mean anything new.  "Sun Grid Engine" is
Codine, which was DQS, and is a cluster resource allocation package.