On Thursday, 05/27/2004 at 01:31 EST, James Melin <[EMAIL PROTECTED]> wrote: > I've done some testing..... and I don't see a speed improvement for FTP of > a large file via FTP to a hipersocket address vs a GBE ethernet connection. > > I did an ftp to the IP address of a Linux guest, and then repeated the send > of that file to the same guest via the hipersocket defined for it. The > results were the same, or slightly FASTER for the GBE. I didn't do enough > iterations to get a clean statistical sample of each method, but it sure > looks to me like it is the routing. > > If you have say, a hipersocket device defined for Linux guest 1 at > 192.168.252.11 and another at 192.168.252.13, that also have a IP address > for a GBE OSA, at 137.70.100.183 and 137.70.100.185 respectively how do you > test so you're connecting from hsi1 on Linux guest 1 with address of 252.11 > to the hipersocket defined as hsi1 on Linux guest 2 at 252.13?
ifconfig will show you the TX and RX packets. If the numbers change, you're using that interface for something. Make sure you are using an MFS of 64K and an MTU of 56K on the HiperSocket if you want to see a speed difference. Of course, if your production data transfers aren't large enough to occupy 56K and keep the buffers full (as FTP does), then the speed differential will fall off accordingly. For small amounts of data, turning the internet crank to get the data on and then off the wire consumes more resource that the actual transmission, so the transmission medium speed is not a big element in the equation. The fewer times you turn the crank, the more important it becomes. Alan Altmark Sr. Software Engineer IBM z/VM Development ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
