Versant's object-level caching allows us to retain only the set of objects used by an application, without the artificial overhead imposed by page-based approaches,'' said Benjamin Treynor, Vice President of Database Technology at Versant. ''The technology covered in the patent takes advantage of Versant's object-level caching to selectively retain those objects most likely to be reused by the application, resulting in a dramatic decrease in cache misses. In a
latency-dominated computing environment, this offers substantial performance improvements when compared to traditional LRU-, CLOCK-, or
frequency-based approaches.''
(From the Versant press release)
Just in case some of the non-techies don't see what this is all about, all the bitchy stuff about "page-based approaches" is a dig at ObjectStore.
Moving on to my interpretation, ODIS has long held a patent on how their caching works, and as those of us who use ObjectStore know, it works incredibly well. Even if Versant offered some credible benchmark showing that their product worked better in a diverse range of plausible real-world scenarios (which I don't think they can or will), ObjectStore is so good anyhow that any improvement would not be terribly significant to customers.
Finally, no wonder Versant is going downhill if they are only interested in sniping at ODIS. What a waste of time. The name of the game is to increase the ODBMS market 100-fold, not to fight against the exitsing players. |
