I think this is an appropriate place to place the SETI crew announcement from their Technical News page as they wrote a great thank you to all who made it possible, from their staff, the School staff and including Ken who wrote SETIQ as well as the users that opened their queques to the public. Outstanding efforts everyone on a worldwide scale again
As you read the news below you can clearly feel it was exciting.
Dave
Network Bandwidth Problems Solved
May 31, 2002
Yesterday at about 23:30 GMT, we switched our data server's route to the internet from the general UC Berkeley link to Cogent Communications with the hope of solving our long standing bandwidth problem. See the Planetary Society's article for a great description of the problem and the Cogent solution.
It is now twenty-two hours after the switch and things look very stable. We are sending on the order of 25 Mb/s outbound through the new link and should have plenty of room to grow. We are no longer dropping connections and the server is operating very efficiently. This ought to translate into successful and speedy data downloads for all of our users.
Many, many thanks go to the networking folks who put this together. The entire route from Space Sciences Lab to Cogent's Palo Alto point of presence was designed and implemented by UC Berkeley's Communication and Network Services (CNS). In particular, we want to thank Siegrid Rickenbach who was the CNS technical lead for this project and got the path between campus and Cogent up and running. Jay Bryon of CNS configured and installed the equipment that was used to trunk both the Cogent and regular campus traffic over the fiber-optic link between campus and SSL. He had to come in very early to schedule outages so that a minimum number of users would be affected. Our own Space Sciences Lab network manager, Greg Paschall, did all of the local work needed for our server to get to the new link. Kudos also go to CNS gurus Michael Sinatra and Ken Lindahl for their routing expertise and critical moment wizardry. Because we had to move off the the SSL firewall, we needed a firewall of our own. Netscreen stepped up with a firewall donation and great support besides.
Thanks also to Ken Reneris who wrote the SetiQueue software package, and to the countless volunteers who used SetiQueue to set up workunit caches, allowing others to get workunits when our server was unreachable. And, of course, thanks to all our patient users who stuck with the project despite the frequent lack of connectivity!
All big changes have unforseen problems and yesterday was no exception. An early morning test revealed a potential show stopper (or at least a show postponer). A test client from offsite, hardwired to come in over the new route, was not able to establish a connection. After thinking a bit at the white board, Michael at CNS figured it out. Incoming connection initiation was happening over the Cogent link. But the outgoing response was happening over our server's default route which was still through campus. This route passed through the SSL firewall which noticed a response to a request that it never saw (the SSL firewall does not see the Cogent link) and denied passage. Well, no problem we thought. We had to change our server's default route at the point of cutover anyway, which would send the response back out over the Cogent link. But not so fast. Our clients utilize the Domain Name Service (DNS) to find our server. DNS still had the old IP address (and by extension the old route). Thus, incoming packets would come in over the campus link and outgoing packets would go out over the Cogent link. The firewall would still see these "half open connections" and deny them. As long as there was asymmetry between the inbound and outbound traffic, we would have a problem. Now, we needed to tell DNS about our new IP address anyway and once this change propagated throughout the Internet we would be OK. But this propagation would take some time, during which nobody would be able to connect. CNS came to the rescue by telling the campus routers to redirect all incoming SETI@home packets to the Cogent link. This change was made at the same moment we changed our default route to send all outgoing packets over the Cogent Link. Symmetry was achieved! The link came up and performed very well. Watching the traffic graphs was an exciting and dramatic moment. We had a speaker phone connection with CNS during the cutover and hoorays could be heard on both ends of the line.
You can see the switchover on both sides: watch the Space Lab bandwidth plummet or the Cogent bandwidth jump.