On Thu, 7 Jun 2012, Edward Ned Harvey wrote:
>> From: [email protected] [mailto:[email protected]] On >> Behalf Of Daniel Feenberg >> >> It is x-windows (over ssh) and http that generate the user complaints, but >> the heaviest load is from long file transfers (gigabytes). > > Do you have an abnormally large (non-default) MTU? Or perhaps, a large no, standard packets only. > parallelization of the aforementioned file transfers? Of course, small > packets can only fit in between large packets, and the routers are not going sometimes we have seen pcs with "download accelerators" that hog all the bandwidth, but shouldn't they cause trouble for pings too? We are looking now for offenders, but can't find any yet. > to interrupt the middle of a large packet to make priority for the small > packet. Even if you have traffic shaping, if you have an abnormally large > MTU, that would still slowdown the small packets. > > Also - This seems obvious, I don't know why I didn't mention it before - > > The problem might be sporadic. You say you measure 20-30ms roundtrip ping > response, but do you do that *all* day long? Do you do that at the same > exact moment that people demonstrate the slowness happening? > We aren't continuous pinging, so it is possible that some intervals are terrible and others ok. I just got back results from the netalyzer. It shows latency 17ms Losss 0.0% tcp connection setup latency 15ms network bandwidth upload 580Kbit/sec, download 1.8Mbit/sec network buffer measurements uplink is good downlink is good So I guess the problem is that X is just too sensitive and we have to implement a replacement. However, we do have a couple of RDP users and they are complaining too. So more bandwidth is on order and we will have to muddle through till it arrives. dan feenberg nber _______________________________________________ bblisa mailing list [email protected] http://www.bblisa.org/mailman/listinfo/bblisa
