John,
I meant that the hardware disk cache is backed by a battery against a power outage. I did not mean a 'solid state disk'.
This paper describes the advantages of a battery-backed cache in a RAID controller:
http://h200001.www2.hp.com/bc/docs/support/SupportManual/c00257513/c00257513.pdf
Looks like 'high-end' RAID controllers, costing from $500 up, typically have battery backing:
http://www.pc.ibm.com/ww/eserver/xseries/scsi_raid.html
Thus, the commit latency in high-end RAID storage really is very small. Thousands of durable commits per seconds are possible.
---
The InnoDB speed of 200 messages per second sounds a bit low. If I understood the program correctly, each message only requires a couple of SQL statements for processing. On a 3 GHz Pentium, you should be able to run about 10 000 simple SQL statements per second. Thus, a 4-processor server might be able to process that 10 000 messages per second if the program is structured in a way where wait latencies are minimized. For example, you may try a busy loop in the GET method.
If different queues are logically independent, why not simply run them in different mysqld servers? That would give unlimited scalability.
Best regards,
Heikki
Oracle Corp./Innobase Oy
InnoDB - transactions, row level locking, and foreign keys for MySQL
InnoDB Hot Backup - a hot backup tool for InnoDB which also backs up MyISAM tables
http://www.innodb.com/order.php