|
Thanks Vince,
This sounds excellent. For me, I can wait, and will
continue using Castor to build a distributed J2EE application. I think Castor's
design is the most correct/simple model.
On the other hand, caching too much affects
performance, as the caches must keep consistent. Without thinking about
reputition, maybe 1 JDO instance to handle a few related database
tables, and the other JDO instances to handle other set of related tables is a
better choice.
I'm looking forward to see your work, great
:)
Thomas
----- Original Message -----
Sent: Monday, July 15, 2002 10:35
PM
Subject: Re: [castor-dev] Performance
cache in a clustered environment
Thomas,
I have a modified version of castor that
provides the ability to expire a
single object, or a class of objects, from
the performance cache. I have
attached a javadoc page that contains a
description of the method
"expireCache" that I added to class "Database"
that exposes this
functionality. This is a modified version of a
similar enhancement I posted
awhile back. This enhancement has been
tested using the base set of tests
provided by the castor test framework,
in addition to two new test cases I
developed specifically to test this
functionality. I have also tested it in
our internal development
environment under load.
I have not submitted this enhancement for
integration into the castor
project since we are not yet using this code in
a production environment. I
would prefer to wait until it has been
proven in a production environment
before submitting it for integration
into the project. But, I will provide
the code and/or a modified set
of 0.9.3.19 jars, to whomever requests a
copy. Please let me know if
you are interested.
Vince
-----Original
Message-----
From: Thomas Phan [mailto:[EMAIL PROTECTED]]
Sent:
Monday, July 15, 2002 12:57 AM
To: [EMAIL PROTECTED]
Subject:
[castor-dev] Performance cache in a clustered
environment
Hi,
> Got the below from a previous post on
Thu, 16 Aug 2001 01:37:32 -0700 by
Tim Fox:
>
> We're
considering building clustering on top of castor using JMS - just in
the
ideas stage at the moment but goes something like the following:
> 1.
Object is inserted/update/deleted in the db (Castor cache is
write-through
so all changes committed in db at end of transaction)
> 2. Send a
message via JMS topic - with the object class and id
> 3. When message
arrives at other instances of Castor - invalidate object
from cache so it
is reread when next requested.
> This should be quite easy to
set-up
>
> Only problem is there doesn't seem to be a Castor
method that invalidates
an object from the cache as far as I know (anyone
please correct me if I'm
wrong) - see (one of my) previous mails.
> I
guess you could explicitly load the object using an OQL query
(apparently
this always bypasses the cache and goes straight to the db) -
although I
don't see how this would deal with the case when the object has
been
deleted - (since the object won't be in the results of the query in
that
case).
> So the question is.. does such a method exist??
I wonder
about the clustering feature in Castor? Will something like the
above be
implemented, or has been implemented? I personally think that the
above
design is great for a distributed J2EE application. It'd be cool to
use
Castor with JBoss 3 Cluster. Otherwise, we will consider OJB, or
EJB2
Thanks
Thomas
-----------------------------------------------------------
If you wish to unsubscribe from this mailing, send mail
to
[EMAIL PROTECTED] with a subject of:
unsubscribe
castor-dev
|
