On Mon, 16 Jun 2008 15:14:52 -0400, François Paré <[EMAIL PROTECTED]> wrote:
>... >I'm running a batch job that does a FTP transfer from a z/OS mainframe to a LINUX server and the transfer rate is about 20K/sec. If I do the same >... I thought I had responded to this but I guess it went into the bit bucket. We had a quite different situation, but it may shed some light on yours. We went through 2 datacenter moves over the past couple years. Each time there was a set of remote FTP clients and servers that had been local but were now a couple thousand miles away. And each time there was a small percent of servers some servers (I'm pretty sure they were always servers) transmission times became terrible. Transmissions to/from similar servers were not much effected by the increased distance. In each case the problem was with the server's TCP/IP windowing scheme - whomever set up the server had not enabled window "scaling". The maximum allowed number of unacknowledged bytes was far too small to allow efficient transmission over a fast media with high latency - a "long fat pipe" (the official term). If you have 2 servers that should be performing similarly but the exchanges with one are very slow, make sure the TCP/IP windowing has been set up correctly. Make sure that window scaling has been enabled. BTW, this is not anything recent. It is described in RFC 1323 from 1992. Pat O'Keefe ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
