Clustering (for Google Summer Code Project)

[Nicolas Toper <nt...@gmail.com>]

Hello,

I am a student in engineering school in computer science (at CNAM in
Paris www.cnam.fr). I work part time as a project lead in a software
company selling ASP services for newsletters. I would be interested in
working on clustering for James.

However before writing a proposal I have a few questions: 

1/ What should be the goal of the clustering: high performance and/or
fault tolerance?

2/ There is already james-ha (not tested). Do you know about this
software? What is good? What is not enough about it?

3/ What should be clustered: James server or its components? I mean we
could either modify the components in James or develop a new layer
upon James. We will do probably a little bit of both but what is the
best way according to you?

Thanks a lot for your answer and I am looking forward to hear from you
Nicolas

---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org

RE: Clustering (for Google Summer Code Project)

[Mike Heath <mh...@avaloninc.com>]

On Sat, 2005-06-04 at 16:25 -0400, Noel J. Bergman wrote:
> First, it isn't part of JAMES, and this is core functionality that we have
> wanted to add (discussion on the list goes back several years, but it hasn't
> been of the highest priority).  Second, james-ha is under the LGPL, which is
> not an acceptable license.

I've relicensed everything in james-ha to be ASL 2.0.  James-ha also
uses jcluster, which is BSD licensed, to wrap JGroups (or any other
group communication framework you want to write a jcluster plug-in for.)

-Mike


---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org

RE: Clustering (for Google Summer Code Project)

["Noel J. Bergman" <no...@devtech.com>]

Nicolas Toper wrote:

> I would be interested in working on clustering for James.

> 1/ What should be the goal of the clustering: high performance
> and/or fault tolerance?

Yes.  And another benefit would be that only the protocol handlers need to
be started as root.  The rest of JAMES could be run as a regular user.

> There is already james-ha (not tested). Do you know about this
> software? What is good? What is not enough about it?

First, it isn't part of JAMES, and this is core functionality that we have
wanted to add (discussion on the list goes back several years, but it hasn't
been of the highest priority).  Second, james-ha is under the LGPL, which is
not an acceptable license.

Likewise, JGroups cannot be used.  ActiveMQ would be an acceptable aolution,
if performance and functionality are up to the task.

> What should be clustered: James server or its components? I mean we
> could either modify the components in James or develop a new layer
> upon James. We will do probably a little bit of both but what is the
> best way according to you?

At the simplest, I would see:

  - Allow JAMES to run as it is, today, all services in a single process.
  - Allow JAMES to run separate services in multiple processes.
    - One or more SMTP Server processes, possibly on multiple machines.
    - One or more pipline processes, possibly on multiple machines.
    - One or more POP3 Server processes, possibly on multiple machines.

The first two are connected by the spool, the last reads from the stores.

Does that help?

	--- Noel


---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org