Home
| Calendar
| Mail Lists
| List Archives
| Desktop SIG
| Hardware Hacking SIG
Wiki | Flickr | PicasaWeb | Video | Maps & Directions | Installfests | Keysignings Linux Cafe | Meeting Notes | Linux Links | Bling | About BLU |
On Friday 10 January 2003 08:04 pm, Bill Horne wrote: > > times). I think I'll build the 100Mb network (stage one) now while > > acccumulating the 64 nodes (stage two) then focus on moving to Gigabit > > (in stage three). I hope this makes sense as I'm still kinda new to this! > > Any thoughts on who I might tap for gigabit donorships/sponsors? > > I read an article about someone that did a cluster using multiple 10/100 > cards per machine, to save on the costs. Try a Google search. I once read an article on this, but I'll never find it. I think it was in Linux Magazine. They put three network cards in each of around 80 machines, and the machines were wired up to three different switches, and somehow the switches were also connected together, and it meant that most machines could talk to each other in one hop, and the rest could do it in two hops. If you lay it out so that the SAN/NFS/cluster controller are accessible to all machines in one hop, though, for most applications it should cover almost all your net traffic needs. I could be misremembering, but I think they had the SAN on one switch and the cluster controller on another switch (OpenPBS?) so that traffic was independent, and the third switch was to the intranet outside the cluster for issuing commands, etc. -- DDDD David Kramer david at thekramers.net http://thekramers.net DK KD DKK D Secret hacker rule #11: hackers read manuals. DK KD DDDD
BLU is a member of BostonUserGroups | |
We also thank MIT for the use of their facilities. |