M5.1 Anza Event

Here is a quick run-down of the web traffic generated by the M5.1 event centered near Anza on the evening of October 30, 2001. The location is in a sparsely populated area, but the event was widely felt, as the Community Internet Intensity Map shows.

Here are graphs of 24 hours of web traffic for four sites in the Earthquake Hazards Program. First is the main program page at earthquake.usgs.gov. The traffic peaked at about 47 hits/sec. No big surprises here. The Program site is the newest of the four sites, so it is not as well-known as the others.
Second is the National Earthquake Information Center page at neic.usgs.gov. Its peak was about 94 hits/sec. This is consistent with normal patterns.

Third is the Pasadena Field Office page at pasadena.wr.usgs.gov. The Pasadena page ordinarily would get a large surge from an event like this, but the server was down that evening, due to the failure of the main network switch in the building where it is housed. The switch failed about two hours before the earthquake, so the server was unavailable for the whole night. The uptick in traffic at 7:30AM shows when the server became available again. Note that the big peak tops out at 130 hits/sec and falls off quickly. This is because the Akamai servers were not able to get fresh content, so there was nothing to serve. Still, the 130/sec peak is probably a good representation of how many people turned to the Pasadena site first.

Finally, we have the main Western Region web page at quake.wr.usgs.gov. The quake.wr.usgs.gov page is probably the best-known source for California earthquake information, so it had the highest traffic, peaking at about 420 hits/sec. This is consistent with past patterns, where about 2/3 of people in Southern California turn first to the Menlo Park site for information.

Note that this peak of 420 hits/sec is now the largest single surge that the Menlo Park site has handled successfully. The original server setup there had a maximum capacity of about 150-180 hits/sec, which would have been insufficient. The Squid server setup could have handled this surge, but might not have been able to handle a larger event. The Menlo Park site was observed to be as responsive as ever right after this event, so it appears that the Akamai EdgeSuite service performed as advertised.