TCP/IP Analyzer RWIN sometimes wrong.

johnhc · Post by **johnhc** » Sun Jul 15, 2007 3:41 pm

I have been using an RWIN of 140160 for some time but sometimes when I run TCP/IP ANalyzer it reports 137240. I have captured a Wireshark trace and the hex equivalent of 137240 scaled by a factor of 4 is not present. The value 137240 is also not present in my Registry. The funny thing is that I used the 137240 value a week or so ago but have rebooted many times since.
Where is TCP/IP ANalyzer getting the erroneous value? Is some router along the way altering the value?
Thanks......John

trogers · Post by **trogers** » Sun Jul 15, 2007 11:50 pm

Do this NDT test and post the report under the 'More Details' button:

http://nitro.ucsc.edu/

Also, post your TCP Analyzer report to compare.

johnhc · Post by **johnhc** » Mon Jul 16, 2007 11:42 am

Thanks, trogers for your response.
I have been running my tests on a separate machine which I don't currently have in service. It will be a few days before I again am using that machine, so I cannot respond to your request at this time.
I do have some observations:
I have used several NDT sites (though not the one you sent) and have observed the 'Details' report. It shows Last, Max, and Min RWIN Received. These numbers almost always differ.
The TCP/IP ANalyzer results are random. If I go to the site, observe the results then click refresh several times, I will the see the RWIN change. Most of the time it will be 140160 (my TCP Optimizer setting) but occasionally it will be 137240. Like I reported, Wireshark shows that 137240 is never sent. I assume that some router along the path is changing the value.
Let me know if you would like any more information such as the 'Details' report for the NDT site and I will send it in a few days.
Thanks..........John

trogers · Post by **trogers** » Mon Jul 16, 2007 11:48 am

When using TCP Optimizer, try setting Max.Duplicate ACKs to 3 instead of the default 2 and see if you get a more stable RWIN value.

johnhc · Post by **johnhc** » Mon Jul 16, 2007 3:03 pm

Thanks, trogers. I'll give it a try when I get back on the machine I am testing on in a few days.....John

johnhc · Post by **johnhc** » Thu Jul 19, 2007 9:24 pm

I set Max Duplicate ACKs to 3 and ran a test with the NSF NDT site. I see little if any difference with the speed results. The minimum RWIN is actually less than it was with a setting of 2. My results using TCP/IP ANalyzer are also about the same. Most of the time my RWIN is reported as 140160 (my setting) while occasionally it is reported as 137240. Again I assume it is being changed en route to the TCP/IP ANalyzer site. I have reset the Max Duplicate ACKs to 2. I am happy with my tweaking results. I have doubled my DL speed on the Swiss NDT site and not hurt the NSF site much. My speeds are typically within the suggested 90%. My DL speed at the large RTT site (e.g. Swiss NDT) is double what it was before increasing my RWIN from the default. I have also unchecked all but the TCP/IP items in my LAN connection properties. I saw no discernible difference and have left them unchecked. My surfing seems to be much snappier.
I have questions remaining about how some on this stuff works but I am very satisfied at my results.
Thanks to all.......John

johnhc · Post by **johnhc** » Wed Jul 25, 2007 3:14 pm

I think I have discovered why my TCP/IP ANalyzer results were sometimes random. It was the way I created the shortcut to the site. I simply dragged the URL from the IE address bar to my links bar. This URL contains the results of the test I had just executed. Apparently under certain conditions (unknown to me) these old results get presented. I am now using http://www.speedguide.net/analyzer.php which seems to give consistent results.
In my search for an answer, I discovered the link in your sticky 'SG Tweaking programs, tools and links ' contains a reference to port 8117 for the analyzer, which was changed to 8080 over a year ago.
Enjoy…..John