Message-Id: <36B06D14.9FDBBBFE@dolphinics.com> Date: Thu, 28 Jan 1999 13:58:44 +0000 From: James Cownie Organization: Dolphin Interconnect Solutions Mime-Version: 1.0 Newsgroups: comp.parallel.pvm,comp.parallel.mpi Subject: Re: Switching HUB for beowulf cluster? References: <78osd7$9ji$1@news.kren.nm.kr> <78piue$lrv$1@pegasus.csx.cam.ac.uk> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Xref: ukc comp.parallel.pvm:7940 comp.parallel.mpi:4562 Nick Maclaren wrote: > > In article <78osd7$9ji$1@news.kren.nm.kr>, Sohn Jung-woo writes: > |> > |> Can somebody recommend me any switching hub for a Beowulf cluster? > |> My laboratory plans to build a beowulf cluster with 10 ~ 16 pentium II > |> servers. While shuffling through some documentations on parallel > |> computing, I found that switching hub plays an important role > |> since network latency is the limiting factor for a beowulf cluster. > > |> 2. How much is the need for 100Mbps bandwidth for a beowulf cluster? > |> Will 10M bps only switching hub be ok for clustering or > |> will 100M bps switching hub lead to better overall performance? > > In terms of latency, there is no difference. In terms of bandwidth, > a lot. As the extra expense of 100M bps is now small, I advise > going for that. > The reason for having a switched hub, rather than a simple repeater is to increase the bi-sectional bandwidth. As Nick rightly says it won't affect the latency at a measurable level, and it also won't improve the point-to-point bandwidth. However because it will allow many concurrent point-to-point messages (unlike a simple hub, or single piece of co-ax) it can affect the application performance immensely. I'd get a 100Mb switched hub, and then worry about latencies in the software a lot. -- Jim James Cownie Etnus, Inc. +44 117 9071438 http://www.etnus.com