10:15 a.m.
Discussing the internals behind data containers in Infinispan.
http://infinispan.blogspot.com/2009/05/implementing-performant-thread-saf...
Enjoy!
--
Manik Surtani
manik(a)jboss.org
Lead, Infinispan
Lead, JBoss Cache
http://www.infinispan.org
http://www.jbosscache.org
3:28 a.m.
New subject: What's behind the data containers: implementing a performant, thread-safe ordered data container
Interesting read. It's a concurrent hash-map of sorts, but one that
can be efficiently traversed by insertion order?
One thing that was a bit confusing, is when you seemingly mention
segment locking in the context of a lock-free data structure. Perhaps
reads are lock-free but writes are protected by a striped locking
scheme?
Anyways, it's nice to witness this attension to concurrent performance
and to read about an interesting datastructure :)
On Tue, May 12, 2009 at 5:15 PM, Manik Surtani <manik(a)jboss.org> wrote:
Discussing the internals behind data containers in Infinispan.
http://infinispan.blogspot.com/2009/05/implementing-performant-thread-safe.html
Enjoy!
--
Manik Surtani
manik(a)jboss.org
Lead, Infinispan
Lead, JBoss Cache
http://www.infinispan.org
http://www.jbosscache.org
_______________________________________________
infinispan-dev mailing list
infinispan-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/infinispan-dev
--
Venlig hilsen / Kind regards,
Christian Vest Hansen.
4:29 a.m.
On 13 May 2009, at 09:28, Christian Vest Hansen wrote:
Interesting read. It's a concurrent hash-map of sorts, but one
that
can be efficiently traversed by insertion order?
Yes, although I don't support all of ConcurrentMap's functionality
(don't need it in DataContainer).
One thing that was a bit confusing, is when you seemingly mention
segment locking in the context of a lock-free data structure. Perhaps
reads are lock-free but writes are protected by a striped locking
scheme?
I don't mention that the data container is lock-free. I do specify
that the linked list is lock-free though. The segment lock when there
is a put() does not protect links. It only protects the LinkedEntry
in question. But if the previous and next entries happen to reside in
different segments, they could be concurrently updated (inserted/
removed) and that would affect linking. And hence the need for a
linked list algorithm that is thread safe and lock-free. Lock-free,
not just for performance but also to prevent deadlocks with the
segment locks.
Anyways, it's nice to witness this attension to concurrent
performance
and to read about an interesting datastructure :)
Thanks!
On Tue, May 12, 2009 at 5:15 PM, Manik Surtani <manik(a)jboss.org>
wrote:
> Discussing the internals behind data containers in Infinispan.
>
>
> http://infinispan.blogspot.com/2009/05/implementing-performant-thread-saf...
>
> Enjoy!
>
> --
> Manik Surtani
> manik(a)jboss.org
> Lead, Infinispan
> Lead, JBoss Cache
> http://www.infinispan.org
> http://www.jbosscache.org
>
>
>
>
> _______________________________________________
> infinispan-dev mailing list
> infinispan-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/infinispan-dev
>
--
Venlig hilsen / Kind regards,
Christian Vest Hansen.
--
Manik Surtani
manik(a)jboss.org
Lead, Infinispan
Lead, JBoss Cache
http://www.infinispan.org
http://www.jbosscache.org
5650
days inactive
5651
days old
infinispan-dev@lists.jboss.org
2 comments
2 participants
participants (2)
-
Christian Vest Hansen -
Manik Surtani