The key element on server load is the Scheduler Daemon... the part you see via the Result Status page, maintaining all the different phases a result is in. The more results that are live on the system, the more time it takes to do the transaction and inform various other parts of the BOINC system of the change. For comparison, the HCC production went to exceed 300,000 daily in the last month from 215,000 average in March. At April's peak of HFCC, this project did 86,000 average.

Overall daily system stats are:

March: 662,268 per day
April: 669,524 per day
May: 780,436 per day (MTD)

WCG routinely held 4 days or more of valid results on the system. For Clean Water, a zero redundant project, that has been shrunk to less than 2 days i.e. Similarly, my oldest valid CEP2 listed, quorum 2, is of Sunday night, less than 4 days old, so it's obvious that WCG techs are pushing harder to reduce the ''live'' number of results on the DB.

Of course this is not to say that storage volume has no impact, but it's by comparison quite subordinate to the number of records held. As the developers say, a Scheduler transaction is the most expensive part [in load terms] of all of the various BOINC modules, which is why the client is designed to minimize the number and tries to combine such things as clearing a Ready to Report with an actual work fetch. The newest client [IN DEVELOPMENT] is btw designed to clear Ready to Report immediately if work fetch has been suspended. Helps those who plan to go on a break from crunching at WCG :D

--//--

Fyi - a single batch from the new target is being sent out to test:
http://www.worldcommunitygrid.org/forums/wcg/viewthread_thread,31574

Seippel