« Broadband capacity - the impact of averaging many subscribers | Main | Broadband capacity - sizing buffers to handle traffic bursts »

December 16, 2009


Guido  Appenzeller

Nice post! One thing to add is that increased latency often also affects throughput as many connections have a limited maximum TCP window size. If the TCP window size is limited, the througput scales with 1/RTT. Thus double latency (as shown in the example above) will reduce throughput by half. ISP's sometimes make use of this by using shapers that intentionally delay ACKs of TCP flows to limit flow throughut without increasing the loss rate.

I don't know how common it is today for flows to be limited by their maximum window size. 5 years ago it was the case for essentially all residential TCP connections, today I would still expect it to be common.

Juliet Waugh

This is really a very informative post.I think we be thankful to you because you have surely done a good job by letting us know this.I liked the idea few programs operate character-by-character, so the extra 100 ms or so might not matter. On the other hand, a once second delay is noticeable!

The comments to this entry are closed.

My Photo

Search this Blog

Subscribe by Email

March 2014

Sun Mon Tue Wed Thu Fri Sat
2 3 4 5 6 7 8
9 10 11 12 13 14 15
16 17 18 19 20 21 22
23 24 25 26 27 28 29
30 31          


Site Meter

Upcoming Travel & Conferences

Twitter Feed

Become a Fan