> In article <m2btj2awyc.fsf@blinky.bfr.co.il>,
>
> > In fact, what seems to be happening is that eventually it's settling
> > on a congestion window size of 1, waiting for each ack before sending
> > another packet. This basically agrees with what I observed after it
> > sped up again, but still doesn't explain the period of time where it
> > only was sending about 1 packet per second. It also doesn't explain
> > why it didn't increase the congestion window during this.
>
> Slow start waits for an ack before doubling the cwnd again, but it still
> retransmits old packets without ack - otherwise it could easily deadlock.
> You were seeing the retransmits. Only when new acks are received the
> sender starts to fill the pipe again.
Ok, so lets say for the sake of argument a packet got nuked & so my
machine retransmitted at about 1 packet/second until it finally got
the appropriate ACKs.
The problem is that when it speeded up again it never completely
filled the pipe. It doesn't look like the congestion window grew as
much as it should have. This was for uploading 3 files of sizes 300k
- 900k, upload times of 2-5 minutes per file. Once it got the
appropriate ACKs & started speeding up again it should have opened the
congestion window & filled the pipe, but this doesn't seem to have
happened.
The other weird thing is that it was very consistent for each file
upload - starts off fast, stalls down to ~1 packet/second & then
picked up again, but never really up to full speed. Why would each
upload suddenly stall after about the same amount of time?
BTW, I just did a couple more tests. It pretty consistently flies for
~20 seconds, then stalls @ 2 packets/second for ~20 seconds & then
picks up again, but never reaching the initial speed.
I still don't see why a) it would consistently stall after 20 seconds
& b) it wouldn't ever recover back to full speed.
-- Harvey J. Stein BFM Financial Research hjstein@bfr.co.il- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/