Re: Relieving some ClassLoader contention in PreferredClassProvider

On Wed, Jan 03, 2007 at 02:11:19PM +0100, Mark Brouwer wrote:
> Hi Peter,
>
> Peter Jones wrote:
>
>> And I think that the challenge with using ConcurrentHashMap here is in
>> correctly dealing with all of the weak references, in both the keys
>> and the values (see above problem).  For comparison, I might suggest
>> looking at the static "localDescs" table of java.io.ObjectStreamClass:
>> prior to JDK 6, it was a regular synchronized HashMap that used a
>> custom internal "EntryFuture" class to accomplish (A); for JDK 6, it
>> was changed to a ConcurrentHashMap for this RFE (which was encouraged
>> by Doug Lea):
>>
>>         http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=5056445

I forgot to point out that this change was backported to 5.0u7 as
well.

> Very interesting! I can't find any details in the release notes of Java
> SE 6 that indicates what kind of performance gains can be achieved. I
> realize there is no official benchmark, but just an indication of the
> range of performance gains with Jini ERI would be very nice.

My recollection is that the responsible engineer (not me, I was a code
reviewer) ran a custom multithreaded serialization benchmark (the
normal internal serialization benchmarks are single-threaded) to
verify that the change provided the same or better performance in a
variety of configurations.  I can't quote numbers, but I recall that
the difference (improvement) varied widely depending on the platform,
number and kind of CPUs, etc.  I don't think that the impact on RMI
(JRMP or JERI) was explicitly measured.

-- Peter

--------------------------------------------------------------------------
Getting Started:     http://www.jini.org/wiki/Category:Getting_Started
Community Web Site:  http://jini.org
jini-users Archive:  http://archives.java.sun.com/archives/jini-users.html
Unsubscribing:       email "signoff JINI-USERS"  to [email protected]