Gigabit copper networking and Cisco gear
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Pete Wyckoff pw at osc.eduMon Oct 21 06:47:12 PDT 2002
- Previous message: Gigabit copper networking and Cisco gear
- Next message: Diagnostic tools
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Justin.Baker at csiro.au said: > The most suitable Cisco switches for a cluster of this size look like > the 10+2 port 3550 or the 8 port 3508G. [...] > Can anyone give me an idea why these switches don't seem to be used > for clusters? I just happen to have recently measured the basic parameters on a few switches found lying around here. The 12-port fiber gigabit ethernet Cisco 3550 is lousy by comparison to the Extreme 6816 or even ancient Packet Engines repeater. http://www.osc.edu/~pw/emp/switch.html Be aware that other factors such as bisection bandwidth and feature set may be important as well. I just happen to care primarily about layer 2 forwarding delay at the moment. And, yes, better throughput numbers can be obtained by playing around with host TCP settings. These measurements just show switch differences (1500-byte MTU and factory default configuration). Foundry and Force10 have dense gigE at perhaps reasonable prices but I have not seen performance numbers. This in case the "cost" argument wasn't enough to convince you. :) -- Pete
- Previous message: Gigabit copper networking and Cisco gear
- Next message: Diagnostic tools
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
