[Rrs-dev] Major speed improvement found

Jim C. Nasby decibel at rrs.decibel.org
Sun Mar 6 22:41:24 GMT 2005


In working on the next version of RRS, I've made a change to the
aggregation query that the system uses to pull raw data into RRS. The
old version of this query would often sequential scan the source table
even though an index scan would be much faster. This change eliminates
that behavior.

The changes I've made for the next version appear to be stable, so
anynoe who wants to pick up the speed improvement should be fine just
using http://svn.rrs.decibel.org/viewcvs.cgi/trunk.tar.gz?view=tar. This
new version also has throttling, that takes effect when update hasn't
run in more than 5x the 'desired run time', which defaults to 50
seconds. 'desired run time' should be set a bit less than your crontab
frequency. If rrs.update() is run every minute, 50 seconds is a good
value to use. If it were set to 60 seconds, many times runs would exceed
60 seconds by a small amount, meaning that the next run would either
exit immediately (if contrib/userlock is installed), or possibly error
out with a uniqueness violation.
-- 
Jim C. Nasby, Database Consultant               decibel at decibel.org 
Give your computer some brain candy! www.distributed.net Team #1828

Windows: "Where do you want to go today?"
Linux: "Where do you want to go tomorrow?"
FreeBSD: "Are you guys coming, or what?"


More information about the rrs-dev mailing list