[ https://issues.jboss.org/browse/ISPN-1300?page=com.atlassian.jira.plugin.... ] Robert Stupp updated ISPN-1300: ------------------------------- Affects Version/s: 5.0.0.CR8 Description: The current implementation of FileCacheStore creates one file for each hash key - which results in up to 4.2 billion files (2^32). It should limit the number of files to a) improve performance of purge b) reduce number of open file handles (system resources) c) reduce number of Java objects/heap (JVM resources) d) improve performance The implementation allows us to do so. Only 4 lines of code are necessary in FileCacheStore implementation: private int hashKeyMask = 0xfffffc00; // TODO should get a configuration entry @Override protected Integer getLockFromKey(Object key) { return Integer.valueOf(key.hashCode() & hashKeyMask); } This reduces the number of files to 2^22 = 4,194,304 files. Since each application and each cache store has different semantics the hasKeyMask value should be configurable - best would be to configure the number of bits. Side effect: If someone changes the FileCacheStore hasKeyMask, the whole cache store becomes unuseable. So I opened another enhancement ... Note: This implementation should be used in a different class (e.g. extend FileCacheStore) because it makes existing file cache stores unusable. was: The current implementation of FileCacheStore creates one file for each hash key - which results in up to 4.2 billion files (2^32). It should limit the number of files to a) improve performance of purge b) reduce number of open file handles (system resources) c) reduce number of Java objects/heap (JVM resources) d) improve performance The implementation allows us to do so. Only 4 lines of code are necessary in FileCacheStore implementation: private int hashKeyMask = 0xfffffc00; // TODO should get a configuration entry @Override protected Integer getLockFromKey(Object key) { return Integer.valueOf(key.hashCode() & hashKeyMask); } This reduces the number of files to 2^22 = 4,194,304 files. Since each application and each cache store has different semantics the hasKeyMask value should be configurable - best would be to configure the number of bits. Side effect: If someone changes the FileCacheStore hasKeyMask, the whole cache store becomes unuseable. So I opened another enhancement ... Component/s: Configuration Loaders and Stores -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira

[ https://issues.jboss.org/browse/ISPN-1300?page=com.atlassian.jira.plugin.... ] Manik Surtani updated ISPN-1300: -------------------------------- Summary: Reduce number of locks and buckets generated by BucketBasedCacheStore (was: Reduce number of files generated by FileCacheStore) Description: The current implementation of FileCacheStore creates one bucket and lock for each hash key - which results in up to 4.2 billion files (2^32). It should limit the number of files to a) improve performance of purge b) reduce number of open file handles (system resources) c) reduce number of Java objects/heap (JVM resources) d) improve performance The implementation allows us to do so. Only 4 lines of code are necessary in BucketBasedCacheStore implementation: {code} private int hashKeyMask = 0xfffffc00; // TODO should get a configuration entry @Override protected Integer getLockFromKey(Object key) { return Integer.valueOf(key.hashCode() & hashKeyMask); } {code} This reduces the number of files to 2^22 = 4,194,304 files. Since each application and each cache store has different semantics the hasKeyMask value should be configurable - best would be to configure the number of bits. Side effect: If someone changes the FileCacheStore hasKeyMask, the whole cache store becomes unuseable. So I opened another enhancement ... Note: This implementation should be used in a different class (e.g. extend FileCacheStore) because it makes existing file cache stores unusable. was: The current implementation of FileCacheStore creates one file for each hash key - which results in up to 4.2 billion files (2^32). It should limit the number of files to a) improve performance of purge b) reduce number of open file handles (system resources) c) reduce number of Java objects/heap (JVM resources) d) improve performance The implementation allows us to do so. Only 4 lines of code are necessary in FileCacheStore implementation: private int hashKeyMask = 0xfffffc00; // TODO should get a configuration entry @Override protected Integer getLockFromKey(Object key) { return Integer.valueOf(key.hashCode() & hashKeyMask); } This reduces the number of files to 2^22 = 4,194,304 files. Since each application and each cache store has different semantics the hasKeyMask value should be configurable - best would be to configure the number of bits. Side effect: If someone changes the FileCacheStore hasKeyMask, the whole cache store becomes unuseable. So I opened another enhancement ... Note: This implementation should be used in a different class (e.g. extend FileCacheStore) because it makes existing file cache stores unusable. Complexity: Low -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira

[ https://issues.jboss.org/browse/ISPN-1300?page=com.atlassian.jira.plugin.... ] Manik Surtani reopened ISPN-1300: --------------------------------- -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira

[ https://issues.jboss.org/browse/ISPN-1300?page=com.atlassian.jira.plugin.... ] Mircea Markus commented on ISPN-1300: ------------------------------------- integrated, thanks for the fix! -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira