In the AsymmetricThroughput case, all indications point to an issue at EsaEstecSite. But we don't know much about that site, except that they have a 1 Gb/s connection to SURFnet. The hosts are not accessible to us, except via FTP and possibly iperf. According to traceroute, the hosts seem to be connected directly to EsaEstecSite's router with the SURFnet uplink. But there may be "transparent" things in between such as rate shapers or firewalls. We should contact Mr. Monleon at ESTEC and try to find out what is there.

On the EvlbiMtuTest, we should move towards closing the case. Most telescope sites now seem to reach 512 Mb/s to the central site in Dwingeloo, with the possible exception of Onsala. Large packets (MTU) also seem to work everywhere.

-- SimonLeinen - 19 Jan 2005

(As MarcoMarletta has said elsewhere, I do not think we should close this case until we have had some more consistent iperf results. That said, in the future we should endeavour to make PERT cases as precisely described as possible, so that they do have a very obvious ending - a second PERT case can always be opened for subsequent work. -- TobyRodwell 19 Jan 2005)

I have checked with Niels from SURFnet to confirm that the link between SURFnet and ESTEC is not a bottleneck. It is not rate limited in any way but there are other customers connected via this link. There is a Gigabit switch inbetween where the organisations are connected. However the whole traffic on the link to the switch inbetween currently peaks at 45Mbit/s (5min average), and the traffic from the Gigabit Switch to ESA is less than 10Mbit/s (5min average). The error counters on the GE switch are all 0, so it is reasonable to believe that this link is not a bottleneck.

I also contacted Mr. Monléon at ESTEC asking about the network infrastructure, but I have not received a reply yet.

-- ChrisWelti - 19 Jan 2005