To: Subramaniam Vincent cc: end2end-interest@ISI.EDU Subject: Re: RED with drop from front Date: Wed, 11 Mar 1998 16:30:34 PST From: Sally Floyd Subramaniam - >Has anybody experimented with RED and "drop from front/head" strategy ? I haven't experimented with drop from front with RED, but I don't think it would be a problem. (I don't think it would be a big win, particularly since with RED the average queue size is usually small. But I don't think it would hurt, either.) When operating RED in byte mode, and taking into account the size of a packet in bytes when deciding the probability with which that packet should be dropped, you have to make sure that you are looking at the size in bytes of the packet that you are actually considering dropping. But other than that, I assume that it would all be completely straightforward. - Sally ------------------------------------------- Subject: Re: RED Date: Thu, 10 Jul 97 14:51:51 PDT From: Sally Floyd ... >Is there any particular reason why RED drops the incoming packet instead of >the packet at the head of the queue? Mostly that I haven't thought about it much. I have assumed that it was easier for routers to drop arriving packets than to drop packets at the head of the queue. But if all packets were the same size, it would be fairly easy to drop packets at the head of the queue, I think. When all of the packets are not the same size, it might be a little tricky. You would have to make sure you weren't measuring arriving packets, and dropping packets at the head of the queue, and inappropriately discriminating against packets of some size. But mainly, when RED is working right the average queue size should be small, and it shouldn't make too much different one way or another whether you drop a packet at the front of the queue or at the tail, I think. The RED congestion indication is not an indication of imminent buffer overflow, but part of some regular steady-state indication to the various end nodes to reduce their arriving rates at appropriate staggered times. - Sally