Mailing-List: contact qpid-users-help@incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: qpid-users@incubator.apache.org
Received-SPF: pass (athena.apache.org: domain of aconway@redhat.com designates
 66.187.233.31 as permitted sender)
Message-ID: <4741C810.4000605@redhat.com>
Date: Mon, 19 Nov 2007 12:29:52 -0500
From: Alan Conway <aconway@redhat.com>
Organization: Red Hat
User-Agent: Thunderbird 2.0.0.9 (X11/20071115)
MIME-Version: 1.0
To: qpid-users@incubator.apache.org,
        Dan Mihai Dumitriu <dan.mihai.dumitriu@gmail.com>
Subject: Re: [Fwd: Re: QPID clustering]
References: <4740F630.1030408@redhat.com>
In-Reply-To: <4740F630.1030408@redhat.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

> -------- Original Message --------
> Subject: 	Re: QPID clustering
> Date: 	Sun, 18 Nov 2007 20:50:22 -0500
> From: 	Dan Mihai Dumitriu <dan.mihai.dumitriu@gmail.com>
> Reply-To: 	qpid-users@incubator.apache.org
> To: 	qpid-users@incubator.apache.org
> References: 	<000901c82a03$3c936420$b5ba2c60$@com> 
> <129a14790711181147y7c07abaen94c44e5635fa81fc@mail.gmail.com>
> 
> 
> 
> I'm actually rather interested in the HA features as well.  Could you  
> expound a bit on how to set up HA using Veritas?  Is it essentially  
> pushing the HA storage problem down to the storage layer (HA DB or NAS)?
> 

I've started some HA work on the C++ broker trunk, its currently in 
suspended animation while I focus attention on performance of the core 
functionality.

So far the cluster supports replicated "wiring" (queue & exchange 
declarations) via OpenAIS, a reliable multicast protocol.

Next step is to replicate queue state. This could be done via multicast 
also but there's an argument that replicating point-to-point to a 
limited number of backups will be make less inclined to saturate the 
network. I expect I will experiment with both to see how it works out. A 
full N-way, all-multicast solution has the advantage of being more 
robust and a lot simpler to configure, but the performance benefits may 
make a limited-backup point-to-point option viable as well.

Thre's some more detail at 
http://cwiki.apache.org/qpid/cluster-design-note.html - its slightly out 
of date as it still talks in terms of the 0-8 protocol, but the concepts 
are pretty much the same.

I expect to get back to this full time within the next month or so, very 
happy to hear any suggestions you might have.

Cheers,
Alan.