...

This is a significant improvement in terms of average RTT, packet loss, throughput and IPDV . However improvement in Ping Unreachability is lesser than expected.

Suggestions from improvement:

Increased Link Capacity:

Reduced average RTT is a consequence of increased link capacity.

Unresolved Issues:

PERN PoP Islamabad  to UETTAXILA

Considering that monitor.niit.edu.pk is not a pern node, we tried to analyze the performance of UET Taxila from Pern PoP Islamabad node.

Looking at the pingtable.pl following trend was observed for various metrics.

  Image Added

High RTT values were observed previously with frequent fluctuations however, RTT values lowered and smoothed out later on. This fact is also supported by the IPDV values before and after the network change as you can see in the graph above. An increment in Throughput values is also evident with the reduction in Packet Loss. All these are results of increased capacity of the network.

This can also be shown in numeric figures by looking at the following table obtained by averaging the pingtable.pl values. The Average RTT reduced by a factor of 1.91, packet loss by a factor of 3, throughput increased 3.14 times and IPDV reduced by a factor of 38.

Find spreadsheet here.

Due to some unavoidable reasons, no smokeping graphs could be generated so far. Once we resolve the issue, we'll show smokeping graphs and traceroutes to support the improvement in network performance. 

Traceroute were taken from PERN POP Islamabad to Uettaxila showing the RTT at each hop.

Executing exec(traceroute, -m 30 -q 3 -f 3, 121.52.150.164, 140)
traceroute to 121.52.150.164 (121.52.150.164), 30 hops max, 140 byte packets
 4  rwp44.pie.net.pk (221.120.254.5)  44.905 ms rwp44.pie.net.pk (221.120.251.29)  24.137 ms rwp44.pie.net.pk (221.120.254.5)  44.803 ms
 5  rwp44.pie.net.pk (221.120.254.61)  45.150 ms  45.167 ms rwp44.pie.net.pk (221.120.254.13)  46.383 ms
 6  rwp44.pie.net.pk (221.120.253.122)  45.073 ms  45.118 ms  45.168 ms
 7  uettaxila.pie.net.pk (221.120.235.126)  58.911 ms  58.896 ms  58.880 ms
 8   (121.52.150.164)  46.431 ms !X  46.175 ms !X  46.519 ms !X

It is observed that the route goes through the 3 interfaces named rwp44.pie.net.pk. Delay to reach rwp44.pie.net.pk is much longer which results in increased RTT for uettaxila. However, from there it takes only 13.8 ms to reach uettaxila.

Suggestions from improvement:

Increased Link Capacity:

Reduced average RTT is a consequence of increased link capacity.

Unresolved Issue:

• Since Friday 5/13/11,
• From March 17th to April 15th there is no data collected from monitor.niit.edu.pk to pinger.uettaxila.edu.pk. Was there some holiday at UET Taxila that resulted in network shutdown?
• Dr. Anjum we carried out all the analysis on seecs (monitor.niit.edu.pk) to UETTAXILA (pinger.uettaxila.edu.pk) link. We all know that SEECS node is not on PERN. In that case is this analysis going to be acceptable to HEC? We haven't chose anyother node to analyze uettaxila because we do not have traceroutes from nodes other than SEECS. Also SEECS nodes are the most reliable nodes as Dr. Les says. (This reliability appears in smokeping graphs having lesser outages from SEECS than from other nodes.)
• From 25th April to 7th May, the Avg_RTT was lowered from 30ms to 5ms and resumed afterwards. I along with Dr. Les tried to figure out the reason but we still are unable to find one. Sir Anjum, Can you suggest any important clue that can lead us to the reason why Avg_RTT was lowered so much?
• Since Friday I am unable to load SEECS pingtable and the smokeping graphs from SEECS pingtable. I need them for analysis. Can someone get this resolved so that I can make report as soon as possible.