Setting innodb_flush_log_at_trx_commit = 2 improves my application performance considerably on my local development environment. The response time and number of requests per second on a 40 concurrent users benchmark is improved by a factor of 8, at least.
However, on my testing environment at Amazon RDS, which is a clone of my production environment, absolutely nothing changes. My benchmark results from after and before setting it to 2 are virtually the same. The server was rebooted and the right value is there when querying for SHOW VARIABLES.
I expected improvement wouldn't be as much as with a local server, but I expected something at all, even if not worth the consequences. Any ideas on what might be wrong? Does Amazon RDS doesn't update it immediately somehow or loads default settings from somewhere else?
