Log in

No account? Create an account
16 July 2008 @ 08:35 pm
For those of you who know anything about the Transport layer, am I doing this right?  (There may be errors in the question phrasing, written by the prof.)

Question:  Consider sending a large file from a host to another over a TCP connection that has no loss.  Suppose TCP uses AIMD for its congestion control without slow start.  Assuming CongWin increases by 1 MSS every time a batch of ACKs is received and assuming approximately constant RTTs, how long does it take from CongWin to increase from 1 MSS to 6 MSS?

Answer:  At the end of 5 RTT's, the CongWin will be 6 MSS.

That seems awfully straightforward to me, but there are these odd little niggling questions... "without slow start" is significant.  I'm presuming that "with slow start" would mean the CongWin doubles every RTT, as it does in examples.

Another small problem I'm having:  the book depicts an RTT as the time it takes the packet to get from the sender to the receiver, but, uh, that's got to be an error... I mean, dude.  ROUND TRIP TIME.  Not one way time.  Right?


* * *

Edit: OGOD. The next question utilized Dijkstra's Algorithm. Kill me. Please. (Can we say "meaningless grunt work", little children? I knew you could!)
Current Mood: confusedconfused
Plaidplaidomatic on July 17th, 2008 02:32 am (UTC)
Is there any prior provided information? Such as: Are we acking every packet, or are we deferring? I ask because the question references "batch of acks". If the receiver is immediately ack'ing on every packet, then yes, 5 RTTs.

There's a lot of otherwise useless or redundant information in the question. The whole "AIMD ... without slow start" is pointless, since you get everything you need from the next sentence.

Also, yes RTT is indeed the full round trip time. Not the one way time.

Wiseacreewin on July 17th, 2008 02:36 am (UTC)
ACKing every packet. Hoping like hell that doesn't present additional delay to be computed. ;)