[e2e] a few questions on TCP mechanism

[Joe Touch]

Michael Welzl wrote:

>>>b. If the only source of feedback is the destination
>>>point, how is it possible for me to:
>>>1. adjust my window size in T < RTT
>>
>>In general, you shouldn't.  As I understand control theory, if you adjust
>>in times smaller than the RTT, your control loop is unstable.
> 
> Great that you bring this up: it's something that I never really
> understood - how does this thinking translate to TCP, which
> updates the rate with every ACK it gets (a lot more often than
> once per RTT)? Is TCP actually interpolating?

TCP doesn't have a 'rate'. It merely keeps track of how many outstanding 
packets it can tolerate. Although the window as a whole is updated for 
each ACK, each ACK updates only a portion of the window that is "one RTT 
old".

TCP adjusts its RTT estimate as well, but that's an average.

On the whole the amount TCP changes per packet is basically limited by 
how much a packet that's one RTT old can affect things.

Joe
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 254 bytes
Desc: OpenPGP digital signature
Url : http://www.postel.org/pipermail/end2end-interest/attachments/20040921/611e8736/signature.bin