Re: [tcpm] Tuning TCP parameters for the 21st century

[Jerry Chu <[e-mail]>]

Hi Erik,

It's good to hear from you!

On Tue, Jul 14, 2009 at 2:46 AM, Erik Nordmark<[e-mail]> wrote:
> Jerry Chu wrote:
>
>> From our own measurement of world wide RTT distribution to Google servers
>> we believe 3secs is too conservative, and like to propose it to be reduced
>> to 1sec.
>
>>
>>
>> Why does it matter?
>>
>> We have seen SYN-ACK retransmission rates upto a few percentage points to
>> some of our servers. We also have indirect data showing the SYN (client
>> side) retransmission to be non-negligible (~1.42% worldwide). At a rate >
>> 1% a large RTO value can have a significant negative impact on the average
>> end2end latency, hence the user experience. This is especially true for
>> short connections, including much of the web traffic.
>
> For short web traffic I'd assume many of the connections go to a server with
> which the client has recently had a tcp connection.
>
> What would be the pros and cons of using the cached rtt measurements from
> previous connections for the SYN?

Yes this is one of the things we've experiemented with. How effective the cache
will depend on how much cache hit and the numbers for us were not very good.
Also Linux dst cache has some implementation/performance issues.

In order to increase the effectiveness of the cache, we're experiementing with a
cache based on /24 (or some other prefix length) rather than
individual dest IP addr.

Note that most likely whatever scheme that may work for the server
side won't work
for the client side but we'd like to see the client side (SYN) to
speed up as well,
hence the proposal to reduce initRTO.

Hope to see you at Stockholm!

Jerry

>
>   Erik
>