Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

The commodity Internet path is longer in terms of no of hops than the TEIN3 path, but slightly shorter in terms of latency, mainly because they go through Mumbai.

Namal has seen similar differences in latency to Japan, Europe etc - no of hops is lower, but the latency is about 40ms higher through TEIN3.

Looking at the PingER data from SLAC to Sri Lanka (SLAC monitors www.ifs.ac.lk, www.ac.lk, www.pgia.ac.lk, www.pgis.lk, and www.ruh.ac.lk in Sri Lanka) the minimum Round Trip Time (RTT) from SLAC appears to have increased by about 10% in line with the above. The average RTT has gone up by 5%. It's too (April 3, 2010) early to tell much about the losses.

Given that the new route although longer has better performance I looked at the inter-packet delay variability (IPDV ~ jitter) which should be better if there is more capacity and less congestion. This is borne out, it looks like the IPDV from SLAC to namunu.learn.ac.lk has reduced from about 4-5ms to less than 0.6ms. This is also seen in the Smokeping plot below. In the plot the blue line is the RTT in ms., the grey smudge around the blue line is the jitter,  the background colors indicate the losses  and black means the host was unreachable at the time the measurement was made (every 30 minutes). !namanu.png!The improved jitter (lack of greay shading) is evident starting on April 1st 2009. The loss of ping connectivity starting on March 24th, 2009 is not understood.

Looking at a Smokeping graph for SLAC to www.ruh.ac.lk one can see a step change upward in the RTT and what looks to be lower jitter starting on April 1st. See below, the black line showing host unreachable from SLAC for a short time on April 1st 2010 may be caused by the cut-over. Image Added

There is also PingER data measured from namunu.learn.ac.lk (see for example http://www-wanmon.slac.stanford.edu/cgi-wrap/pingtable.pl?file=average_rtt&by=by-node&size=100&tick=last60days&from=LK.LEARN.AC.N1&to=WORLD&ex=none&only=all&dataset=hep&percentage=any and choose your metric and time ticks). I have not looked at this yet in any details to see what it reveals. I do notice the minimum RTT from LEARN to Australia went up by ~ 100ms, while the IPDV dropped from about 0.8ms to about 0.3ms.