Expensive black boxes (Re: Subject: good, cheap switches?)
erayo at cs.bilkent.edu.tr
Mon May 13 21:15:12 PDT 2002
On Monday 13 May 2002 19:27, Donald B. Kinghorn wrote:
> ... take this for what it's worth ... I've used dozens (hummm, probably
> hundreds) of inexpensive Netgear switches for smallish (<32 node) clusters.
> I've only had one with any problems [ a single bad port] I like the 500
> series like the 24 port 524 . These switches use a "cut through" design
> instead of the usual "store and forward". These are no frills unmanaged
> switches but seem to work just fine for the internal network of a cluster
> and you can get the 524 for around $300
What exactly is the topology/architecture of the switch?
I looked for such info on our 3com superstack ii 3900 switch but I couldn't
find any. Maybe such details are unimportant for computational scientists but
parallel programming research still makes use of the network architecture. I
find it highly immoral that companies try to keep these secret while selling
the devices for several thousands of dollars.
A CT-routing hypercube or mesh, or an omega network would make great sense to
me, but I haven't seen such information on any of the high-end switches.
If possible, I'd prefer a switch that I would know rather than an expensive
black box that claims to be a clique with 0 latency.
If anybody from 3com is reading this consider this a harsh inquiry.
Eray Ozkural (exa) <erayo at cs.bilkent.edu.tr>
Comp. Sci. Dept., Bilkent University, Ankara
www: http://www.cs.bilkent.edu.tr/~erayo Malfunction: http://mp3.com/ariza
GPG public key fingerprint: 360C 852F 88B0 A745 F31B EA0F 7C07 AE16 874D 539C
More information about the Beowulf