SSD Caching performance

Synthetic benchmarks are essentially useless in this sort of testing. We need to use some sort of trace based testing (i.e. PCMark) in order to capture something as close to the userland ‘seat of the pants’ feel. We did our testing with the following configurations:

  • Western Digital Caviar Green 2TB (solo)
  • Western Digital Caviar Green + SSD 311 (enhanced mode)
  • Western Digital Caviar Green + SSD 311 (maximized mode)
  • Intel SSD 311 Series (solo)
  • Intel SSD 320 Series 300GB (solo)

We ran each configuration three sequential times. While we would normally average results, I wanted to demonstrate the dynamic nature of results in the caching configurations. Here are the results:

You will notice the HDD results are (understandably) the lowest of all. The next two (cached) series show HDD-like performance on the first run, but a very rapid increase to close-to-SSD like performance on the second and third runs. We were surprised to see how close the SSD 311 came to the SSD 320, considering the 320 has a significantly higher write speed rating as well as double the internal data pathways to the flash.

Looking at the enhanced vs. maximized mode differences, it looks like enhanced can get you about 90% of the way there. Those wanting the extra 10% can just switch on the fly to maximized, but that carries with it the added risk of having data out of sync between the cache and the HDD. This is not a big deal most of the time, but comes into play when failures and power loss scenarios occur. I’d personally stick with enhanced mode unless I was extremely confident in the stability of my system (and had it connected to a good UPS).

« PreviousNext »