From: jhpark@nurapt.kaist.ac.kr (Jeong Hwan Park) Newsgroups: comp.parallel.mpi Subject: Re: Poor MPI performance with collection of Linux workstations Date: 29 Mar 1999 07:03:15 GMT Organization: KREONet news service Message-Id: References: <7cn7up$4tq$1@grandprime.binc.net> <19990324180108.01062.00001229@ng-fs1.aol.com> Mime-Version: 1.0 Content-Type: text/plain; charset=EUC-KR Content-Transfer-Encoding: 8bit Xref: ukc comp.parallel.mpi:4856 the effective backplane bandwidth of Intel express 510T is 400Mbps. Read "Bay Networks, Inc. BayStack 450-24T Switch Fast Ethernet Switch Competitive Evaluation Backplane Capacity Test Summary" at http://www.tolly.com/online_1/testresults/documents/qdoc.cfm?Doc=232 >>I >>have nine machines, one of which is a 4-processor box. They are all >>200 MHz PPros. I get wildly varying execution times, too. I am running >>an atmospheric numerical model in a 3-d torus architecture through an >>intel switch, 100baseT, Digital DS21140 Tulip ethernet cards (half >>duuplex) in each machine. Each machine (node) runs the same code, >>which goes like (1) calculate stuff (2) exchange boundary data between >>adjacent nodes (3) repeat. > >I suspect the problem is with the switch. I also used an Intel switch (Intel >Express 510T) with the cluster I used for my thesis. Literature says the >switching fabric has an aggregate capacity of 6.3Gbps. But also, on page 78 of >the users manual, it notes that the effective aggregate capacity offered to the >network is 800Mbps. As a result of this, when I had more than eight nodes >communicating at once, the performance was variable and generally poor. At >best, the nodes detect the carrier already on all available timeslots; at >worst, packets are lost due to collisions and the nodes don't realize this >until they time out waiting for ACK. > >Take care, >cb >Christopher A. Bohn EngrBohn@aol.com > http://members.aol.com/EngrBohn/ >"Oooh! What does >this< button do!?" -- ------------------------------------------------- Jeong Hwan Park (jhpark@nurapt.kaist.ac.kr) Korea Advanced Institute of Science and Technology Department of Nuclear Engineering 373-1 Kusong-dong Yusong-gu, Taejon 305-701 South Korea --------------------------------------------------