At the start of June 2014 year the route from SLAC to NUST appeared as below:
The blue lines are from the Maxmind database, the red lines are from Raja’s et. al VTrace.
On 1/17/2015 it was found to go the other way around the world:
The blue lines are from the Maxmind database, the red lines are from Raja’s et. al VTrace.
It appears the change in direction happened between Jun 10 and Jun 11 2014. See the traceroutes below for the 2 days:
Note the 2014-07-11 traceroute that goes West from California is different from todays. The visualization of the 2014-07-11 route appears below:
It appears the change in direction happened between Jun 10 and Jun 11 2014. See the traceroutes below for the 2 days:
Note the 2014-07-11 traceroute that goes West from California is different from todays. The visualization of the 2014-07-11 route appears below:
Explanation
The answer is the HEC's use of its secondary service provider transworld. Transworld is linked to outside world through France, hence the path that was being used before June 2014.
When the internal backbone connection is switched back to the main provider (Partially owned by HEC and partially PTCL), the normal link from Singapore is used. In general, International traffic from Pakistan follows through Singapore route that is now being used. some service providers also use Dubai link, however, I do not think that is the case with for PERN data.
The question is: When HEC uses Transworld link and for which universities? There is no definite answer. Probably because of the upgrade to PERN 2 or load balancing, I am not sure. If you remember, this Transworld link was first discovered when Faisalabad pop had exceptionally high RTT when pinged from within Pakistan. This was documented in October 2012 report and shared with HEC. At that time, Faisalabad pop was on Transworld backbone and because of absence of interISP switching gateway, traffic from other pops to faisalabad was being routed internationally (Karachi to US to UK to France to Faisalabad). Hence very high RTT. When HEC was inquired about it, they simply switched Faisalabad on primary service provider and did not explain why it was put on secondary service provider.
What are the effects of being on Secondary Service Provider?
High RTT values when that specific node is accessed from other nodes within Pakistan.
Different RTT compared to other pakistani nodes when that node is accessed from outside world (for example SLAC). Higher are lower is not a discrete matter. Apparently link capacities are same, Which international nodes gets shorter path when a specific link is used makes the difference.
What can we do?
I believe its not a healthy trend when certain nodes are put on the secondary service provider network and others and the primary network.
It can be a very good research study with good potential of publication. Someone will need to sit down and study all impacts of this link switch. In case of faisalabad pop, node was less frequently used. In case of NUST, its more frequently accessed and hence more information can be drawn from this switch. There must be other nodes in the path there were switched to secondary provider network (for example PERN POP). If someone is interested, I can help through it but given other commitments I might not be able to contribute too much. It can for sure be a masters thesis.
Should we report it to HEC? That is a decision I will leave to Dr. Arshad. Generally, informing them does not make a difference. They sometimes take it positive, at other times, they take it negatively and fire back