You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@openjpa.apache.org by Craig L Russell <Cr...@Sun.COM> on 2007/05/17 06:21:52 UTC
Fwd: Lock-free HashMap
Don't know if anyone caught Cliff's preso at Java One...
Anyone want to take a look at the new HashMap?
Craig
Begin forwarded message:
> From: Kin-man Chung
> Date: May 15, 2007 11:24:38 AM PDT
> Subject: Lock-free HashMap
>
> Cliff Click talked about his lock-free HashMap at JavaOne, and
> demonstrated its superior scalability over ConcurrentHashMap.
> The source is available at
>
> http://sourceforge.net/projects/high-scale-lib
>
> We should check this out and see if it can be used in Glassfish to
> boost the spec numbers.
>
> -Kin-man
>
Craig Russell
Architect, Sun Java Enterprise System http://java.sun.com/products/jdo
408 276-5638 mailto:Craig.Russell@sun.com
P.S. A good JDO? O, Gasp!
Re: Fwd: Lock-free HashMap
Posted by "David Ezzio (asmtp)" <de...@bea.com>.
Hi Craig,
Another ConcurrentHashMap implementation that is available for JDK 1.4
comes from Emory University's Distributed Computing Lab.
(http://harness2.org/util/backport-util-concurrent/dist/) One of their
goals to make the source easily upgraded to java.util.concurrent once
JDK 1.4 is left behind.
Given the sneakiness of concurrency, I would be very leary of using
anything that hasn't been thoroughly vetted.
I think we can have high confidence in java.util.concurrent as well as
get a good performance boost. Its only downside is that it is not
available for JDK 1.4.
David
Craig L Russell wrote:
> Don't know if anyone caught Cliff's preso at Java One...
>
> Anyone want to take a look at the new HashMap?
>
> Craig
>
> Begin forwarded message:
>
>> From: Kin-man Chung
>> Date: May 15, 2007 11:24:38 AM PDT
>> Subject: Lock-free HashMap
>>
>> Cliff Click talked about his lock-free HashMap at JavaOne, and
>> demonstrated its superior scalability over ConcurrentHashMap.
>> The source is available at
>>
>> http://sourceforge.net/projects/high-scale-lib
>>
>> We should check this out and see if it can be used in Glassfish to
>> boost the spec numbers.
>>
>> -Kin-man
>>
>
> Craig Russell
> Architect, Sun Java Enterprise System http://java.sun.com/products/jdo
> 408 276-5638 mailto:Craig.Russell@sun.com
> P.S. A good JDO? O, Gasp!
>
Re: Fwd: Lock-free HashMap
Posted by Knut Anders Hatlen <Kn...@Sun.COM>.
David Van Couvering <da...@vancouvering.com> writes:
> Something that may be of interest to Derby as well...
Thanks for the link, David! It looks interesting. Derby already uses
ConcurrentHashMap in the lock manager, so it should be relatively easy
to plug in a lock-free HashMap instead, I think. I'll see if I can find
time to test it out.
What would be really interesting would be if we could rewrite more of
the globally shared HashMaps in Derby (buffer manager and transaction
table come to mind) to use ConcurrentMap and provide a mechanism to
enable users to choose which implementation to use. By default, Derby
would use ConcurrentHashMap which is part of the JDK, and with a
property the users could specify that they wanted Cliff's
implementation (or someone else's).
> ---------- Forwarded message ----------
> From: Craig L Russell <Cr...@sun.com>
> Date: May 16, 2007 9:21 PM
> Subject: Fwd: Lock-free HashMap
> To: open-jpa-dev@incubator.apache.org
>
>
> Don't know if anyone caught Cliff's preso at Java One...
>
> Anyone want to take a look at the new HashMap?
>
> Craig
>
> Begin forwarded message:
>
>> From: Kin-man Chung
>> Date: May 15, 2007 11:24:38 AM PDT
>> Subject: Lock-free HashMap
>>
>> Cliff Click talked about his lock-free HashMap at JavaOne, and
>> demonstrated its superior scalability over ConcurrentHashMap.
>> The source is available at
>>
>> http://sourceforge.net/projects/high-scale-lib
>>
>> We should check this out and see if it can be used in Glassfish to
>> boost the spec numbers.
>>
>> -Kin-man
>>
>
> Craig Russell
> Architect, Sun Java Enterprise System http://java.sun.com/products/jdo
> 408 276-5638 mailto:Craig.Russell@sun.com
> P.S. A good JDO? O, Gasp!
--
Knut Anders
Fwd: Lock-free HashMap
Posted by David Van Couvering <da...@vancouvering.com>.
Something that may be of interest to Derby as well...
---------- Forwarded message ----------
From: Craig L Russell <Cr...@sun.com>
Date: May 16, 2007 9:21 PM
Subject: Fwd: Lock-free HashMap
To: open-jpa-dev@incubator.apache.org
Don't know if anyone caught Cliff's preso at Java One...
Anyone want to take a look at the new HashMap?
Craig
Begin forwarded message:
> From: Kin-man Chung
> Date: May 15, 2007 11:24:38 AM PDT
> Subject: Lock-free HashMap
>
> Cliff Click talked about his lock-free HashMap at JavaOne, and
> demonstrated its superior scalability over ConcurrentHashMap.
> The source is available at
>
> http://sourceforge.net/projects/high-scale-lib
>
> We should check this out and see if it can be used in Glassfish to
> boost the spec numbers.
>
> -Kin-man
>
Craig Russell
Architect, Sun Java Enterprise System http://java.sun.com/products/jdo
408 276-5638 mailto:Craig.Russell@sun.com
P.S. A good JDO? O, Gasp!
Re: Lock-free HashMap
Posted by David Jencks <da...@yahoo.com>.
On May 17, 2007, at 3:41 PM, Marc Prud'hommeaux wrote:
>
> I recall Patrick was looking into replacing some of our slower Map
> plumbing recently. If he is still working on that, it would be
> interesting to see if these could be dropped in to see what the
> performance differences were.
From the presentation I believe it really helps to have a 400+ core
machine and lotsa threads to see the performance differences, which
are then rather dramatic :-). Also the current implementation uses
some com.sun classes (unsafe??) for CAS instructions and I got the
impression that it uses slightly more memory than ConcurrentHashMap.
Looking forward to the results :-)
thanks
david jencks
>
>
>
>
> On May 16, 2007, at 9:21 PM, Craig L Russell wrote:
>
>> Don't know if anyone caught Cliff's preso at Java One...
>>
>> Anyone want to take a look at the new HashMap?
>>
>> Craig
>>
>> Begin forwarded message:
>>
>>> From: Kin-man Chung
>>> Date: May 15, 2007 11:24:38 AM PDT
>>> Subject: Lock-free HashMap
>>>
>>> Cliff Click talked about his lock-free HashMap at JavaOne, and
>>> demonstrated its superior scalability over ConcurrentHashMap.
>>> The source is available at
>>>
>>> http://sourceforge.net/projects/high-scale-lib
>>>
>>> We should check this out and see if it can be used in Glassfish to
>>> boost the spec numbers.
>>>
>>> -Kin-man
>>>
>>
>> Craig Russell
>> Architect, Sun Java Enterprise System http://java.sun.com/products/
>> jdo
>> 408 276-5638 mailto:Craig.Russell@sun.com
>> P.S. A good JDO? O, Gasp!
>>
>
Re: Lock-free HashMap
Posted by Marc Prud'hommeaux <mp...@apache.org>.
I recall Patrick was looking into replacing some of our slower Map
plumbing recently. If he is still working on that, it would be
interesting to see if these could be dropped in to see what the
performance differences were.
On May 16, 2007, at 9:21 PM, Craig L Russell wrote:
> Don't know if anyone caught Cliff's preso at Java One...
>
> Anyone want to take a look at the new HashMap?
>
> Craig
>
> Begin forwarded message:
>
>> From: Kin-man Chung
>> Date: May 15, 2007 11:24:38 AM PDT
>> Subject: Lock-free HashMap
>>
>> Cliff Click talked about his lock-free HashMap at JavaOne, and
>> demonstrated its superior scalability over ConcurrentHashMap.
>> The source is available at
>>
>> http://sourceforge.net/projects/high-scale-lib
>>
>> We should check this out and see if it can be used in Glassfish to
>> boost the spec numbers.
>>
>> -Kin-man
>>
>
> Craig Russell
> Architect, Sun Java Enterprise System http://java.sun.com/products/jdo
> 408 276-5638 mailto:Craig.Russell@sun.com
> P.S. A good JDO? O, Gasp!
>