Thank you very much.
So actually the redirection is not transparent to the client, nor there
is a "run time" redirection, since the client declares in the HTTP
request the keys it's going to use.
If this last sentence is correct, for the time being I don't have any
further question. Thank you for your explanation.
Regards,
Diego
Responding to this here in case it's interesting to other
members.
I'm using some of the new features introduced in 4.2 to really boost performance.
My distributed cache nodes expose their business logic -- consisting of multiple ISPN
caches -- via REST. There is a very high data contention rate that acquiring a
cluster-wide lock is too expensive. Also, a single business operation touches multiple
caches, each with their own natural keys.
So to get the throughput I needed:
1. I used ISPN's "eagerLockSingleNode" option. There's an explanation
on the Wiki about it.
2. I made sure that the keys across multiple caches would be collocated, even across a
rehash. I had to create a custom key container that would return the same hash code.
There is no guarantee that two keys generated by ISPN's key factory would be on the
same node after a rehash.
3. The DistributionMangerImpl allows you to get the owner of a key -- and whether a key
"isLocal()". So since I'm using HTTP, I can examine the request, get the
keys, and find out who owns them. I can then send a redirect (30x) back to the clients,
and they forward the request to the data owner.
4. Because of the ISPN configuration, I now can start a transaction on the data owner
and acquire only local locks.
This improved throughput in my system considerably -- almost by a 4x factor.
Hope this helps.
Erik
-----Original Message-----
From: infinispan-dev-bounces(a)lists.jboss.org
[mailto:infinispan-dev-bounces@lists.jboss.org] On Behalf Of Diego Didona
Sent: Wednesday, February 23, 2011 5:52 PM
To: infinispan -Dev List
Subject: Re: [infinispan-dev] API stability policy for major vs minor vs patch releases
Hi Erik!
I'm a little bit off-topic but I would like to ask you what do you mean by
"redirection". Can you explain it to me?
Thanks
Diego
> +1.5B
>
> This method is a core component of my local transaction/redirection implementation.
Hopefully, this was an oversight and not intended to be removed.
>
> Erik
>
> -----Original Message-----
> From: infinispan-dev-bounces(a)lists.jboss.org
> [mailto:infinispan-dev-bounces@lists.jboss.org] On Behalf Of Paul
> Ferraro
> Sent: Wednesday, February 23, 2011 12:01 PM
> To: infinispan-dev(a)lists.jboss.org
> Subject: [infinispan-dev] API stability policy for major vs minor vs
> patch releases
>
> In my course of upgrading AS6 from 4.2.0.Final to 4.2.1.CR3, I came across a
NoSuchMethodError. Specifically, a couple AS components utilize
DistributionManager.isLocal(String), which was dropped from the public API sometime after
4.2.1.CR1.
>
> While the fix is trivial enough (in the end I'll need to perform several
component release to compensate), this raises the larger issue of a API stability policy
for major vs minor vs patch releases. At the very least, I don't think its wise to
remove methods from a public interface in a patch release. In my opinion, removing
methods from a public API should only happen across major releases, and even then, only
after being formally deprecated.
>
> Thoughts?
>
> Paul
>
> _______________________________________________
> infinispan-dev mailing list
> infinispan-dev(a)lists.jboss.org
>
https://lists.jboss.org/mailman/listinfo/infinispan-dev
>
> The information contained in this message is legally privileged and confidential, and
is intended for the individual or entity to whom it is addressed (or their designee). If
this message is read by anyone other than the intended recipient, please be advised that
distribution of this message, in any form, is strictly prohibited. If you have received
this message in error, please notify the sender immediately and delete or destroy all
copies of this message.
>
> _______________________________________________
> infinispan-dev mailing list
> infinispan-dev(a)lists.jboss.org
>
https://lists.jboss.org/mailman/listinfo/infinispan-dev
_______________________________________________
infinispan-dev mailing list
infinispan-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/infinispan-dev
The information contained in this message is legally privileged and confidential, and is
intended for the individual or entity to whom it is addressed (or their designee). If this
message is read by anyone other than the intended recipient, please be advised that
distribution of this message, in any form, is strictly prohibited. If you have received
this message in error, please notify the sender immediately and delete or destroy all
copies of this message.
_______________________________________________
infinispan-dev mailing list
infinispan-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/infinispan-dev