The CPU stats for the last week are shown in the graph before
The blue line (i-06bab47e) is the public facing server and the green line (i-b89e41c7) is the second server.
The main server runs the web site, upload & download, validator and assimilators. The second server runs the tasks - which includes keep the queues feed.
Until the 25th we were stressing the second server, but not the main one. Once the challenge started both servers started to get loaded. They survived for a couple of days; then on Sunday morning the disk on the main server crashed up under the load. Fortunately I'd just gotten up to feed the horses (at 5am) and noticed it.
I stopped the servers and moved them from the m1.small to the c1.medium instance type under AWS and restarted them.
As you can see the load on the main server is now quite manageable, whilst the second server is peaking when I have to load the BOINC queues with more galaxies.
International Centre for Radio Astronomy Research