I would say caching, on multiple levels (CPU, DB, File System). By
splitting at least some of the load, it's possible for parts of the
cache to become old and get flushed. When everything is on one
machine, the box has a complete picture of the traffic patterns and
can optimize better.
On Jul 31, 2007, at 8:17 AM, nigel wood wrote:
A puzzler for you guys.. all plausible explanations (and
suggestions for proving them) gratefully received.
We run several MySQL database servers in the traditional master-
slave configuration and attempt (rather poorly) to spread select
queries between them. Normally the slave gets 1/3 of the master
load. Both machines have identical configurations, hardware
specifications and network connectivity. The main clients of these
databases are PHP websites without persistent connections. A fail-
over pair of machines in a separate building replicates from the
master.
Today (as a result of replication failure) we directed all the
traffic normally sent to the reporting server back to the master
server adding a 1/3 to its load. Several areas of the websites got
FASTER afterwards and I'm currenlty at a loss to explain why.
Nigel Wood
--
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe:http://lists.mysql.com/mysql?
[EMAIL PROTECTED]
--
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe:http://lists.mysql.com/[EMAIL PROTECTED]
