On Thu, 13 Mar 2003, Thomas Corell wrote:
After a short look in the archive: AP DB ratio(DB/AP) 100 250 2.5 worked
(but not enough connections to handle traffic, many http connects were dropped at peak activity)
240 250 1.042 errors 175 400 2.286 let's wait
I thinks wikipedia's current scripts will need a 2.25 to 2.5 ratio to work properly.
The current worst-case scenario should be 3.0, if every persistent connection breaks and every available http connection is used by a sysop to run SQL queries (which presently opens a second connection under a separate mysql user). Plus a small constant for manual connections for maintenance or automatic updates of RDF feeds, backups, etc.
Since there are a relatively small number of sysops, and few of them will be running queries simultaneously, we shouldn't need that much reserve space.
-- brion vibber (brion @ pobox.com)