GigaSpaces, in conjunction with MPI Europe, Globant, and Intel recently conducted some benchmarks on the in-memory data caching / data grid element of their version 7 XAP platform on Intel’s Nehalem chipset. XAP’s latest version 7 reached 1 million data updates per second and 2.6 million data retrievals per second with four client threads on the Nehalem Chip.
Previously XAP version 6 had benchmarked 276,000 updates per second and 453,000 retrievals per second on the best previous Intel processor.
The summary of the tests are:
– GigaSpaces Write and Take operation from their in-memory data cache are about 3-4 times faster (300-400%) with the Nehalem chipset (in absolute numbers).
– Read operations perform much better with the Nehalem (3-6 times better with 1-4 threads). As much as there are more concurrent threads the difference is increasing. According to Shay Hassidim, one of the reasons for this is the non lock read capability added to XAP 7.0.
– Nehalem+ XAP 7.0.1 shows better scalability than Dunnington+XAP 6.2.2. About 30 % better with write and take operations and growing numbers with read operations (90% with 10 threads).
GigaSpaces continues to push the speed and performance envelope with its product, and I’m informed that the 7.02 version of GigaSpaces has again been highly performance tuned and is even faster than the 7.0 platform that was used for this benchmark.
It will be interesting to see if other vendors in this space publish results of their product on Nehalem, which looks set to deliver a huge performance jump.