Thanks, that worked!! I still had to use a periodic sleep to allow cleanup to catchup but now I can insert large number of objects (500,000) without running out of heap space.
I still have the issue with the <cachename>.key file being zero'd out at the end of data access (using the same configuration, and calling the dispose method at the end of the program). Now here's the next question. what kind of read performance can I expect using JCS. Using the same cache file created above (500,000 records) , I end up taking about 35 seconds for every 10,000 reads (presumably from disk I/O) . Thats about 285 reads per second . Seems slow to me. Is there anyway of speeding this up, GIVEN that I'm starting from a persistent FILE. Keep in mind that this rate seems to slow down, the larger the number of records in the cache file. Thanks for all the help. On Mon, Jul 21, 2008 at 10:16 AM, Dennis Jacobs <[EMAIL PROTECTED]> wrote: > I'm using JCS 1.2.7.9 to accomplish something similar (I believe). Try > configuring the cache region (memory) with: > jcs.default.cacheattributes.DiskUsagePatternName=UPDATE > > This will force the write out to disk when putting items in the cache, > instead of waiting for a shutdown. In order to get the keys written out as > well I shutdown the cache. I've had some trouble getting it to shutdown > gracefully if it needs to do a lot of cleanup (large disk optimizations or > serialization from the memory cache); by using the UPDATE option to write > to > disk during puts this keeps the shutdown quick and tidy. > > There's probably an api which you could call in code to write out keys as > well without shutting down, but I haven't tried it. Once the keys are > written out you can just copy the .key and .data files to any number of > target machines. > > Best of luck, > > Dennis > >
