This content has been marked as final. Show 5 replies
In that case, are you sure that lowering the read service time will speed up the GG processing of the logs? Just want to be sure that lowering the read service time will indeed speed up the process - no sense tuning it if that's not the bottleneck.
If it is and you think read service time is the only bottleneck at this time, then the only option I know of would be to create a diskgroup out of a portion of the flash disks. This reduces the flash cache size available to the system, so be sure that the benefit will outweigh the cost.
I don't think we have a way to pin online redo logs in flash cache currently. Others can correct me if I'm wrong.
At Oracle OpenWorld the Exadata Smart Flash Log feature was announced. The Smart Flash Log feature requires Exadata Storage 188.8.131.52.0 or later, and Databases version 184.108.40.206 Bundle Patch 11 or greater. This feature allows a modest amount of flash to be used as a secondary write place. It wites redo to both flash and disk and returns the call to the db for the first one that finishes. By doing so it improves user transaction response time, and increases overall database throughput for IO intensive workloads.