We've done this all the time with our caching system in our production
system, using RAM-DISK-REMOTE cache.  The use of RAM along is not good
enough as there is no upper bound to the cache item total size.  The use of
DISK is justtified as the direct access time to the file system is way
faster then the time of querying the db.

Serializable is actually desirable in such cases, to the very contrary.  And
experience shows it works really well.

We also have a hard/soft reference hybrid cache in our remote cache system,
where the soft cache items are dynamically removed according the the JVM
memory need.  (Hard item by definition must stay in memory/disk.)

Hanson

-----Original Message-----
From: Stephen Haberman
To: [EMAIL PROTECTED]
Sent: 7/26/2002 11:41 PM
Subject: possible application for jcs?

Hi,

A Torque user, Scott Eade, has submitted a patch for Torque to deal with
paging through large chunks of database rows. It's got a lot of logic to
deal with what rows to keep in memory and when it needs to go fetch some
more.

I had the thought while reviewing his patch that perhaps JCS could be
used as the store? My hesitation is that it would involve a fair amount
of puts...say getting 5-6 page blocks at once (with, say, 100 rows per
page), putting them all in memory, then waiting a few seconds/minutes
for the client to work through those, then putting another 5-6 pages
blocks in memory, etc.

Would this be too many writes to make JCS effective? For what it's
worth, I couldn't for see using any type of disk/network cache, it'd be
primarily in memory only.

(Perhaps this is a good application where Serializable wouldn't be
desired, but only wanting to keep non-serializable Objects around for so
long).

Thanks,
Stephen


--
To unsubscribe, e-mail:
<mailto:[EMAIL PROTECTED]>
For additional commands, e-mail:
<mailto:[EMAIL PROTECTED]>

Reply via email to