On Tue, 2006-04-25 at 20:12 -0700, Rob wrote: > How can I verify that a 1Gb/s network is indeed > operating at its optimal speed? I tried this: > > [master]$ ping -s 65507 node > 65515 bytes from node: icmp_seq=0 ttl=64 time=1.97 ms > 65515 bytes from node: icmp_seq=1 ttl=64 time=1.95 ms > 65515 bytes from node: icmp_seq=2 ttl=64 time=1.94 ms > 65515 bytes from node: icmp_seq=3 ttl=64 time=1.97 ms > > (I tried many times, over a long period of time to > get these typical values). > >From this I conclude that it takes about 1.95 ms > for 65515 x 8 bits to go forth and back between > master and node. > > Ideally, on a 1Gbit/s network, the time should be: > 65515 x 8 x 2 / (1024^3) = 0.98 ms > (x 2 for the roundtrip signal forth and back > and 1024^3 is the 1G of the network) I wouldn't be using ping to test transfer *speed*. Ping's sending a small packet, and something else responds. That's a *latency* check of your network and the computers either side of it. For a brute force check, you could transfer a large number of bytes from a byte generator (/dev/zero?) to /dev/nul, and time it. Don't try a large file copy, you'd be measuring the speed of your disc drives, if you want to test the speed of your networking. Though, if your concern for how fast your system is, is how long it takes to copy files between boxes, then a test like that is quite practical. -- (Currently running FC4, occasionally trying FC5.) Don't send private replies to my address, the mailbox is ignored. I read messages from the public lists.
