In-memory cache | FlushToDiskOnStop
13.4.3 Offload to disk
By default, the dynamic cache maintains the number of entries that are configured in memory. If new entries are created while the cache is full, the priorities that are configured for each cache entry, and a least recently used algorithm, are used to remove entries from the cache. In addition to having a cache entry removed from memory when the cache is full, you can enable disk offload to have a cache entry copied to the file system (the location is configurable). Later, if that cache entry is requested, dynamic cache first evicts a cache entry from JVM based on priority and LRU algorithm, and then re-inserts the needed cache entry into memory.The disk cache size can be controlled in WebSphere Application Server Version 6.0.2.17 or later.
The disk cache size can be controlled in terms of the number of entries or size of disk. You can do so by setting JVM custom properties...
IBM recommends that the customer use SAN disks for hosting the disk cache for performance-critical real-time applications.
In theory, disk cache is only limited by file system size. However, in practice consider the performance overhead versus performance gain.
Larger disk cache size is dependent on how fast the information can be written to disk and retrieved from it. If you allow disk cache to grow without limit, eventually you will reach a point where it is faster to regenerate a cache entry from the database than to look it up from disk.
The ideal disk cache size must be determined in an endurance test.
A PMI counter, HitsOnDisk, may be used to monitor the number of hits on disk cache.
xxxx