I just wanted to show some simple testing results from using ioTurbine. The environment consists of a HP BL460G7 blades which has a 320GB IO Accelerator from FusionIO. The iocache is used to optimize read operations and provide a level of performance that exceeds the norm of what you’ll typical see in your environment. I did benchmark using iometer using the settings for a 100% read that were 100% random at a 4k block size.
And then on the Disk Targets tab I set the maximum disk size to 1GB or 2000000 sectors and 32 outstanding IOs. And because I have two vcpu on the VM, I configured two worker processes.
After running the tests with and without caching turned on here are the results.
The difference is very clear that the iocache provided greater IOPS and lower IO response times or latency. These benefits can only be acheived if your using the cache thought but we can see from the terminal that the cache hit ratio is greater than 99% during the testing of 100% random reads.