Clustering VS. Mainframe

David Mandala plug-discuss@lists.plug.phoenix.az.us
22 Jan 2003 13:03:48 -0700


A few more reasons that people go to clusters are:

1) Failover/Down time. If a single unit in a cluster dies the rest keep
on working. If the cluster is big enough it may even be hard to notice
that a single unit dies. (Google uses special clusters for their search
engine.)

2) Cost, for many problems that can be split over a cluster it is
usually cheaper to build a big cluster then buy one machine and item one
becomes a factor too.

3) Harness spare CPU cycles. Some folks have lots of workstations and
they run specific software to distribute software compiler or other
types of intensive calculation work such as seti @ home and non
dedicated rendering farms. They gain extra horsepower from machines they
already own for other purposes.

Some scientific clusters do use the exact same hardware and netboot to
avoid issues of maintenance and to make sure the systems are running the
same software. Some of them have been forced to upgrade nodes because
the old unit types are no longer available regardless of price. In that
case the administrators have built a system of netbooting that supplies
boot images that are correct for each type of node. More complex but
still easier then maintaining lots of HD's.

Some mainframes look like large clusters to the software running on
them. The IBM 390 running Linux can have thousands of Linux instances
running on the same machine. The software thinks it's on a cluster but
the actual hardware is the mainframe. This is a special purpose item.
The mainframe of course cost big $$$ but there are cases where this is
cheaper then a cluster of real hardware when you calculate MBTF and
floorspace and head costs.

Cheers,

Davidm

On Fri, 2003-01-17 at 12:33, Liberty Young wrote:
> I was just pondering the following question based on what David said
> earlier today.
> > 
> > Lots of people would rather spring for a cluster then huge single
> > machine now a days.
> > 
> 
> 
> Putting aside that it's very geek-cool to run or put together a cluster
> of cheap x86 pcs, what are the reasons to have a cluster vs a single
> machine? 
> 
> One reason TO have a cluster is that you can run your own clustering OS
> and kernel; then again, the kernel and OS may fall shy of the ones built
> into and for the single machine. 
> 
> I was also thinking that with clusters, it is very hard to amass a clone
> army of the same pc for each node. You would HAVE to have a contract
> with a vendor that states very explicitly that they would supply you an
> exact replica for each node. Otherwise, there'd be difficulty  and costs
> involved in maintance. For example, you couldn't just use a single
> restore CD or a ghost image of the machine to install a new one or fix a
> corrupted one. 
> 
> Disclaimer: These are just theoritcal thoughts. I'm not a large scale
> administrator of any kind. 
> 
> ---------------------------------------------------
> PLUG-discuss mailing list - PLUG-discuss@lists.plug.phoenix.az.us
> To subscribe, unsubscribe, or to change  you mail settings:
> http://lists.PLUG.phoenix.az.us/mailman/listinfo/plug-discuss
-- 
David IS Mandala
gpg fingerprint 8932 E7EF CCF5 1B8C 1B5C A92E C678 795E 45B2 D952
Phoenix, AZ (480) 460-7546 HP, (602) 741-1363 CP
http://www.them.com/~davidm/