-
1. Re: Cache performance with millions of nodes
genman Feb 6, 2007 12:53 PM (in response to jivesociety)You might want to use the Berkeley DB instead of the FileCacheLoader. There's also a JDBM implementation (I wrote :-) ) that might be adequate.
Or, create a "deeper" tree where there are fewer children per node. E.g. if you were to create a phone book, instead of just dividing the entries by "A, B, C" etc., further divide them into "AB" "AC" etc. -
2. Re: Cache performance with millions of nodes
jivesociety Feb 8, 2007 7:04 PM (in response to jivesociety)Does the JDBM implementation support multiple processes writing to a file?
thanks in advance,
ray -
3. Re: Cache performance with millions of nodes
genman Feb 9, 2007 2:15 PM (in response to jivesociety)It does, but there is heavy thread locking done. It's not as fast as this:
http://sleepycat2.inetu.net/products/bdbje.html
You might get decent performance from the JDBM using the async cache loader. -
4. Re: Cache performance with millions of nodes
genman Feb 9, 2007 9:15 PM (in response to jivesociety)I might have misunderstood you. If you have multiple _processes_ (not threads) doing writes to a file, you need some sort of central coordinator, and that would require you to use a traditional RDB. JBCache does support JDBC databases.