Java ehcache disk store

Question

I'm working on application in which I need to generate and frequently access thousands of files. For disk space usage reasons, I only want to keep around a fixed number of these files at any given time. For example, the files are being written to C:\my-folder. Once my-folder reaches 1000 files, if I need to save a new file, I would like to erase the LRU file from my-folder. Is something like this possible using ehcache (or any caching tool)? I thought I could use a disk store in ehcache, but whenever I call get on the cache, its only looking at the keys that are in memory and not in disk.

Some snippets of my code are shown here:

    // create the cache
    CacheManager cm = CacheManager.create();
    String name = getName();
    CacheConfiguration config = new CacheConfiguration(name, 1)
            .maxElementsOnDisk(2000).diskPersistent(true)
            .overflowToDisk(true).eternal(true).diskStorePath(name)
            .memoryStoreEvictionPolicy(MemoryStoreEvictionPolicy.LRU);
    Cache cache = new Cache(config);
    cm.addCache(cache);

    // I do a couple of puts
    cache.put(new Element("key1", val1));
    cache.put(new Element("key2", val2));    
    cache.flush();

    // now key1 is no longer in the cache (since max memory size is 1), but I'd like to look on disk since I have set maxElementsOnDisk to 2000
    Element el = cache.get("key1");

Thanks,

Jeff

Answer 1

This doesn't seem like you are using EH Cache to it's potential.. to store 1000 items on disk and then rotate away the last used one. EH Cache is more for storing stuff in memory, and using disk as a backup when memory fills.

I would personally just roll my own. You have very specific business rules. Store 1000 items. Then when you are ready to store file 1001, you could go and find a file to delete. If you have access time turned on at the OS level, you can do a single linux command to delete the least recently accessed file... Usually you don't use access time for files in linux, but it would make your problem easy to solve at the OS level..