Mailing-List: contact users-help@mina.apache.org; run by ezmlm
Precedence: bulk
Reply-To: users@mina.apache.org
Received-SPF: pass (nike.apache.org: domain of Steve.Ash@metavante.com
 designates 38.114.34.151 as permitted sender)
Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Subject: RE: Questions on MINA as relates to SEDA, and web services
Date: Thu, 11 Dec 2008 10:06:34 -0600
Message-ID: <CDF1367226739B4C964368A30CD2B7F40DDB75D3@prdsvc05.vectorsgi.com>
In-Reply-To: <125748861020762318408602286363734369010-Webmail@me.com>
Thread-Topic: Questions on MINA as relates to SEDA, and web services
Thread-Index: Aclbqeow0RyCNKd7QWO43T8I47W3zwAADExw
References: <125748861020762318408602286363734369010-Webmail@me.com>
From: "Steve Ash" <Steve.Ash@metavante.com>
To: <users@mina.apache.org>

I can't offer any answers, but have many of the same questions you do as
it sounds.=20

Also- I think anyone saying "excellent MINA documentation" must be
speaking sarcastically.  There are a few examples, but I have found the
documentation lacking with regards to  architecture, etc.  Not knocking
MINA- I use it and I like it, but would LOVE more comprehensive
documentation with best practices, etc.

Steve Ash

-----Original Message-----
From: Steve Rehrauer [mailto:foolyap@mac.com]=20
Sent: Thursday, December 11, 2008 9:50 AM
To: users@mina.apache.org
Subject: Questions on MINA as relates to SEDA, and web services

Hi -

I've spent the last couple of years on a team building a WAN server for
an older LAN product.

The server design was fixed before I joined the team (not to cast blame,
just to say that it was
in-flight and deemed immutable).  The implementation uses JEE
technologies (web services via
SOAP over HTTP).  Scalability of this has been disappointing; we
struggle to support 200
simultaneous users.

The client that connects to the server is based on the eclipse platform,
and is rather greedy of
server resources (makes multiple simultaneous requests).  Requests can
take many minutes to
complete; this is due both to the flexible nature of the requests (i.e.,
the client can ask for huge
amounts of data), and to processing that takes place in the LAN
component that is ultimately
being used.  Connections stay open for the duration of these
long-running requests.  Each
request can consume multiple threads in the Java tier.  Each client can
make multiple requests
for a user gesture.

In a brief lull between releases, some of us have been casting about for
alternatives that would
scale better.  In particular, that reduce the number of threads running
in the Java tier.  This has
led us to Java NIO, thence to SEDA, and from there to MINA.

I have some (possibly naive :-) questions about MINA that I hope someone
can answer.

First, does it even seem sensible to use MINA to try to implement an
HTTP protocol?  I've seen
references to the asyncWeb project, but it seems like it's perhaps still
being baked?  Put another
way, is anyone building and shipping a commercial product that uses it?
Or, does anyone know
of another open-sourced HTTP solution built atop of Apache MINA?  (We
may have the freedom
to move away from web services & SOAP, but we'd still be using an
HTTP-based protocol.)

Second, are there any reasonably-size worked examples out there of using
MINA to implement
SEDA concepts?  I see folks referring to the "excellent MINA
documentation", and I confess I
feel stupid as a result. :-)

Have many more questions, but am curious to see responses to the above
first.

Thanks, --Steve