Re:Scalable SQL: How do large-scale sites and applications r...
by rudiger on Friday April 22, 2011 @03:46PM
(#2475198)
Interesting. I've never seen large-scale database deployments use SSDs (neither for the transaction log, nor for the data); I've only seen battery-backup and write-back caches on hard disk drives used.
Re:Scalable SQL: How do large-scale sites and applications r...
by jpitz on Saturday April 23, 2011 @01:58AM
(#2476351)
You will probably start to. The intel 320 series is getting some attention now [1] with a combination of competitive ( for SSD ) price and enough capacitors on board to flush the write cache on a power failure.
You'd likely use SSDs first for indexes, then data. Transaction logs are usually a sequential workload and 7200 rpm drives often suffice.
[1] http://archives.postgresql.org/pgsql-performance/2011-03/msg...
Re:Scalable SQL: How do large-scale sites and applications r...
by zwischenzug on Friday April 22, 2011 @04:11PM
(#2475265)
Nor have I. But it may be the future: cf rethinkdb.
Some tables are sharded, others aren't. Some tables are imported into hadoop for hive queries, others aren't. Backups are run using a fragile program called zrm and an army of outsourced DBAs to manage it.
The army of outsourced DBAs also deal with replication failures, figuring out why inexplicable deadlocks happen, tuning every inane innodb configurable, and "rebuilding" servers when mysql decides to slow itself down and yearns for a complete dump/reload of all tables.
Friends don't let friends use mysql.