Sahil, A TCP Sender will reduce the congestion window in a similar fashion to both a) first ECE bit received in a window (flight of packets worth one RTT), and of course, b) exceeding the DupThresh limit of duplicate acks, again in one RTT worth of segments. The typical reaction in these two cases is a multiplicative reduction of the size of the congestion window, with the muliplication factor 0.5 - with other words, the sending rate is halved. (Some variants of TCP deviate from this factor of 0,5, but it's always a multiplicative decrease (the second half of AIMD - additive increase, multiplicative decrease; the general control law that results in stable network operating under load).
A retransmission timeout causes the sending TCP to reduce it's sending rate much more significantly - after an RTO, it restarts the TCP sending rate much like from a clean start, often even more gentle, as the initial window is reduced to only 1 or 2 segments (instead of 4 to 10)). Regarding the absolute number of losses - note that it's only one loss (or mark) per RTT that halves the sending rate; with large congestion window, you could drop dozends or even hundreds of segments, and the sending TCP will still only half the sending rate. As long as all these losses occur within the first RTT. Another loss after the first RTT, and the TCP sender will react again (thus send only at 1/4 from the original rate). This is the reason why looking at raw drop numbers it not all that helpful byitself, the overall combination of goodput and latency (which, in case of TCP, includes queuing-induced latency, transmission latency and loss recovery latency) are much more relevant. With ECN, the loss recovery latency is virtually zero, as the congestion signal is explicitly signaled from the network to the end hosts. Unfortunately, Latency hasn't been a big topic for ISPs; Datacenter operators care much much more about latency (and it's easier there to deploy solutions, like ECN with an AQM set up with good parameters for that limited environment). Hope this helps, Richard ----- Original Message ----- From: sahil grover To: Jonathan Morton Cc: [email protected] ; Richard Scheffenegger Sent: Friday, February 27, 2015 3:34 PM Subject: Re: [Codel] why RED is not considered as a solution to bufferbloat. Jonathan, crystal clear now.. i mean the way u explain, Hats off seriously. well,i need to learn some more. My question is how tcp reacts to packet drop, like if we talk about packet marking then with acknowledgement from receiver, tcp sender gets the congestion signal(if i am not wrong) but if packet is dropped then what is mechanism? is it related with (a) 3 duplicate acknowledgement or (b)timeout event? On Thu, Feb 26, 2015 at 7:26 PM, Jonathan Morton <[email protected]> wrote: Okay, let me walk you through this. Let's say you are managing a fairly slow link, on the order of 1 megabit. The receiver on the far side of it is running a modern OS and has opened its receive window to a whole megabyte. It would take about 10 seconds to pass a whole receive window through the link. This is a common situation in the real world, by the way. Now, let's suppose that the sender was constrained to a slower send rate, say half a megabit, for reasons unrelated to the network. Maybe it's a streaming video service on a mostly static image, so it only needs to send small delta frames and compressed audio. It has therefore opened the congestion window as wide as it will go, to the limit of the receive window. But then the action starts, there's a complete scene change, and a big burst of data is sent all at once, because the congestion window permits it. So the queue you're managing suddenly has a megabyte of data in it. Unless you do something about it, your induced latency just leapt up to ten seconds, and the VoIP call your user was on at the time will drop out. Now, let's compare the behaviour of two AQMs: Codel and something almost, but not entirely, unlike Codel. Specifically, it does everything that Codel does, but at enqueue instead of dequeue time. Both of them will let the entire burst into the queue. Codel only takes action if the queue remains more full than its threshold for more than 100ms. Since this burst arrives all at once, and the queue stayed nice and empty up until now, neither AQM will decide to mark or drop packets at that moment. Now, packets are slowly delivered across the link. Each one takes about 15ms to deliver. They are answered by acks, and the sender obligingly sends fresh packets to replace them. After about six or seven packets have been delivered, both AQMs will decide to start marking packets (this is an ECN enabled flow). Let's assume that the next packet after this point will be marked. This will be the receiver's first clue that congestion is occurring. For Codel, the next packet available to mark will be the eighth packet. So the receiver learns about the congestion event 120 ms after it began. It will echo the congestion mark back to the sender in the corresponding ack, which will immediately halve the congestion window. It will still take at least ten seconds to drain the queue. For NotCodel, the next available packet for marking is the one a megabyte later than the eighth packet. The receiver won't see that packet until 10120 ms after the congestion event began. So the sender will happily keep the queue ten seconds long for the next ten seconds, instead of backing off straight away. It will take at least TWENTY seconds to drain the queue. Note that even if the link bandwidth was 10 megabits rather than one, with the same receive window size, it would still take one second to drain the queue with Codel and two seconds with NotCodel. This relatively simple traffic situation demonstrates that marking on dequeue rather than enqueue is twice as effective at managing queue length. - Jonathan Morton
_______________________________________________ Codel mailing list [email protected] https://lists.bufferbloat.net/listinfo/codel
