Yep. Most, if not all, lock-free algorithms I have seen rely heavily on
atomic updates (the Atomic* family and volatiles) and spin-locks. While
this is usually a good tradeoff in a local situation, in a distributed
setting it is rather expensive.
Steven Harris wrote:
Probably wouldn't perform well distributed. It does a lot of fine
grained updates which would be expensive over a network. Distributed
algorithms generally
try to do as much locally as possible and minimize cross node
chatter. Might be an interesting experiment though. Then maybe come
up with variations
that improve it.
Cheers,
Steve
On Oct 8, 2007, at 9:47 AM, Fernando Padilla wrote:
Good to hear.
As an aside. I was wondering how the "Lock-free Hash Table" from the
Azul guys behaved in Terracotta? Have you guys heard of it? It just
sounds like a map made for massive multi-thread access, and Terracotta
would go well together.
http://blogs.azulsystems.com/cliff/2007/03/a_nonblocking_h.html
http://72.14.253.104/search?
q=cache:um30fXUqFkgJ:www.azulsystems.com/events/
javaone_2007/2007_LockFreeHash.pdf+non+blocking+hash
+map&hl=en&ct=clnk&cd=1&gl=us&client=firefox-a
Steven Harris wrote:
For those who play with nightly builds their has been some
significant improvement made to clustered ConcurrentHashMap in trunk.
It now takes advantage of read locks and partial values. Please give
it a try and let us know how it works for you.
Cheers,
Steve
_______________________________________________
tc-dev mailing list
[email protected]
http://lists.terracotta.org/mailman/listinfo/tc-dev
_______________________________________________
tc-users mailing list
[EMAIL PROTECTED]
http://lists.terracotta.org/mailman/listinfo/tc-users
_______________________________________________
tc-users mailing list
[EMAIL PROTECTED]
http://lists.terracotta.org/mailman/listinfo/tc-users
_______________________________________________
tc-dev mailing list
[email protected]
http://lists.terracotta.org/mailman/listinfo/tc-dev
