Re: Which scales better? J2EE, PHP, or CGIDEV2?

["Dean, Robert" <rdean@xxxxxxxxxxxx>]

I just checked my source, I had my understanding backwards....

The architecture is that each incoming connection doesn't require a thread.  Threads are only consumed while the request is being serviced.

This is especially important in server push contexts (Comet), where having a thread per connection would have even worse scalability than stateful connections.

-----Original Message-----
From: web400-bounces@xxxxxxxxxxxx [mailto:web400-bounces@xxxxxxxxxxxx] On Behalf Of Aaron Bartell
Sent: Thursday, October 07, 2010 1:42 PM
To: Web Enabling the AS400 / iSeries
Subject: Re: [WEB400] Which scales better? J2EE, PHP, or CGIDEV2?

This is an interesting concept, but how do they do that without
requiring some sort of asynchronous approach where the server
communicates back with the client in a new request? (basically
switching roles)

Does it require special code on the client?  Does it require the
client to poll the server?

Aaron Bartell
http://mowyourlawn.com
http://mowyourlawn.com/blog/



On Thu, Oct 7, 2010 at 12:07 PM, Dean, Robert <rdean@xxxxxxxxxxxx> wrote:
> Yes they do.  Some of the app servers (Jetty for sure) also manage connections the same way i manages jobs (i.e., requests that are waiting for backend I/O are shuffled off to a separate pool and the sockets are reused for other requests).  This allows concurrent requests to scale beyond the number of TCP/IP ports available.