Science Talk

Max Power · Guest

============
By changing the parameter of "Maximum daily WU quota per CPU" from 150 (what
it is today) to something more reasonable like 15, this database problem
could fix itself almost instantly.

When the database issues calm down, only increment "WU per CPU" by 5 and
then see how it impacts database performance. This CPU number should be
moved up and down until all DB backlog abnormalities disappear.

Max Power, CEO
Power Broadcasting
DSN @ Home : http://hireme.geek.nz/dsn-at-home.html
============

2008-03-31 00:20:46 UTC -- Database Problems Due to Sieve Speed

http://boinc.rieselsieve.com/

The massive increase in sieve speed (almost doubling our output) in the last
week has really started to strain our DB. The biggest reason is because
people are queuing a TON of work. We've gone from ~60,000 workunits
outstanding at any point to more than 195,000 outstanding.

I've found more than 100 hosts that are responsible for about 50% of this
queue, each with more than 450 workunits outstanding. Please consider
reducing queue sizes, as it puts a huge strain on the database when hosts
are queuing 1000 or more workunits.

Please expect intermittent downtime through the next few days as I have to
sometimes take the DB down for 15-20 minutes to get more work on board and
optimize the database tables.

Additionally, workunits will start going out that are 2G in length, rather
than 1G, to help cut down on how many WU we're going through.