> An interesting side note: Ethereal apparently doesn't support window > scaling.
Do you mean the tcp window scale option or just the phenomenon of having data outstanding? You can always just count un-acked bytes for an idea of how big the window is. It's dyanmic, of course, and don't forget about slow start, which says that during the start of a TCP connection, the window starts out small and increases to the configured maximum. Many of these parameters are tunable on some OSes, including (to a certain extent) Windows. I only use ethereal these days, and this is what I do. Understanding of Slow Start (and related algorithms) as defined in RFC 2001 is important in order to understand TCP performance charasteristics. For example, at an ISP I did network architecture for a cuople of years ago, we really felt the pinch because of the way HTTP operates - it makes lots of very short-lived connections. This not only causes an extra 3-way handshake and 4-way disconnect each time, but *every* connection has to go through slow start. This results in extremely restricted throughput. In other words, traffic flow was: 1 SYN---------> 2 4 HTTP GET----> 5 8 11 13 FIN---------> 14 I use a download test site that will tell you your rcv window if > it remains consistent enough throughout the test. It follows exactly what I > enter into the registry, whether above or below 65535. However, when using > a value above 65535 (such as 93360), Ethereal reports it only as 65535. > This may have been why I was only "seeing" a range of 64420 - 65535 > earlier. This further confused the situation. Serves me right for using an > open-source PA, but the price is right. > > Thanks all for your insightful feedback. You were a great help. And I was > able to do what I originally set out to - seriously restrict my download > rates by manipulating this value. I'm hopeful it will go a long way towards > demonstrating a very technical issue to somewhat non-technical people. It's > something tangible for them to wrap their minds around sitting right in > front of a desktop machine. > > Regards all, > > Scott -- ---------------------------- Steve Dispensa Chief Technology Officer Positive Networks Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=59440&t=59400 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
