Hi Rob Thanks for sharing your observations. Please could those at Brunel and Sheffield in particular take note of the SACK/DSACK settings at their sites and experiment with these changes too? We can request additional tests when you are ready. Many thanks, Jeremy On 30 Jun 2010, at 12:25, Rob Fay wrote: >> I'd be interested if applying this sort of tuning to worker nodes >> will have any effect at the other sites that are having transfer >> problems - Brunel, Liverpool, Sheffield and Bristol. > > We tried it at Liverpool last Thursday, and sure enough, the quality > of the transfers went into the dark green. > > However, I can't quite see how the optimisations would explain that > particularly given the seeming correlation with NAT. Why would NAT > cause problems just because the transfers are taking longer? > Wouldn't larger transfers, or increased contention, also cause > problems if that was the case? > > So looking at the changes, aside from the changes to buffer sizes, > there are two TCP options turned off, timestamps and selective > acknowledgements (along with duplicate selective acknowledgements). > > I tried turning timestamps back on, that made no difference, > transfers stayed green. > > I then tried turning SACK and DSACK back on and we saw a drop in > quality over the next few hours down into the 30-40% orange/yellow > range. Disabling SACK and DSACK saw the quality go back to 100%. > > I then restored all settings to defaults apart from SACK and DSACK > being disabled, and all transfers since then have been 100%. > However, there haven't been that many transfers since then, so I > don't think I can really say with certainty that SACK/DSACK are the > issue, but the evidence so far would appear to indicate that may be > the case, at Liverpool at least. > > -- > Robert Fay [log in to unmask] > System Administrator office: 220 > High Energy Physics Division tel (int): 43396 > Oliver Lodge Laboratory tel (ext): +44 (0)151 794 3396 > University of Liverpool http://www.liv.ac.uk/physics/hep/