"You don't have to rearchitect or configure your applications in any way, shape or form. You don't have to do anything on backend storage," Flint continued. "Neither end even knows the caching is happening. It sits in the middle, automatically identifies hot, active data, places a copy on high-speed media and the applications by extension go faster."
Flint said on standard databases, the caching software can triple performance. On OLTP applications, which are more read intensive, performance can jump 18-fold, he said.
While application servers perform a certain amount of caching in volatile DRAM already, the amount of caching is limited by the 4GB to 8GB of memory typically on board a server. Intel's CAS software takes advantage of higher-capacity SSDs, which can have as much as a terabyte of capacity, to improve the performance for far more data.
Performance improvement varies on systems and depends on the ratio of backend data to active data on the server and whether the data is read or write intensive. CAS is read-acceleration cache.
"If your application is doing nothing more than a bunch of re-reads, you'll have the same performance on our two-level cache as a single-level SSD cache," Flint said. "On the other hand, in scenarios where you're running databases ... where things like indexes or temp tables become extra hot for a period of time, we can ensure those indices and temp tables are in memory, and in doing so, we can drive performance out of the system that goes well above from what you'd get from just moving all your data to SSD."
To ensure it does not duplicate work already being performed in DRAM, the CAS software takes control of all data access. The hottest data is placed on DRAM, while less hot, but still I/O active data, is placed on SSD.
Intel's CAS software provides cooperative integration between the standard server DRAM cache and the Intel SSD cache, creating a multi-level cache that optimizes the use of system memory and automatically determines the best cache level for active data.