Well, "theory" says that the receive window size is based on how much memory a host has. If it has very little memory, then it has to make sure the other side doesn't send too much data without giving the host breathing time to flush its buffer, write to disk, and ACK. So a host has to advertise a smallish receive window when it's low on memory.
So one wild idea would be to remove memory from the host. Or maybe you could get it to use up a chunk of memory by opening lots of large docs and/or using a RAM disk!? You would think you could just configure it, but that doesn't seem to be working, eh? Good luck. Keep us posted. Thanks. Priscilla s vermill wrote: > > Marc Thach Xuan Ky wrote: > > > > Scott, > > A clue from this webpage: > > http://www.psc.edu/networking/perf_tune.html > > Describing Win98 it says "DefaultRcvWindow is a string type > and > > the > > value describes the default receive windowsize for the TCP > > stack. > > Otherwise the windowsize has to be programmed in apps with > > setsockopt." > > Perhaps the app is setting it differently. It also seems to > > imply that > > GlobalMaxTcpWindowSize should do it since the OS should > enforce > > this on > > the app. Do you know what units the variable uses? that > website > > indicates that the default is a gig, so it may be measured in > K > > or M, > > just a thought. > > Hey...I hadn't thought of that. None of these keys even > existed in my registry when I started out, so it's impossible > to know what the default was. I just assumed it was > octet-based. Let me try some new numbers assuming K or M. > > Thanks! > > > rgds > > Marc > > > > > > s vermill wrote: > > > > > > Marc Thach Xuan Ky wrote: > > > > > > > > Are you trying to make the window smaller? > > > > rgds > > > > Marc > > > > > > Yes. I was hoping to set up a demonstration on the impact > of > > high > > > bandwidth*delay product networks without actually having a > > high > > > bandwidth*delay product network. By artifically enforcing a > > small rcv > > > window, I should get about the same result. > > > > > > Thanks Marc, > > > > > > Scott > > > > > > > > > > > s vermill wrote: > > > > > > > > > > On a W2k machine, I've tried several different > > > > recommendations for adjusting > > > > > the TCP receive window size. None of them, including > > those > > > > directly from > > > > > Microsoft, seem to have any impact. I'm capturing my > own > > > > traffic and my > > > > > advertised window is always in the 64k range. > > > > > > > > > > I've tried editing the \tcpip\parameters to include > > > > 'TcpWindowSize' and > > > > > 'GlobalMaxTcpWindowSize' - neither of which had any > > effect. > > > > I've tried > > > > > editing \VxD\MSTCP to include 'DefaultRcvWindow' - also > no > > > > effect. > > > > > > > > > > Anyone know how to manipulate the rcv window that my > > machine > > > > will > > > > > advertise. For that matter, what about the other MS > OSes? > > > > XP? Win98? > > > > > > > > > > Thanks all, > > > > > > > > > > Scott > > > > > > Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=59429&t=59400 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
