Re: EGL vs. Java

Justin Taylor wrote:

Joe:
The reason I ask about the speeds you are quoting is that we
have a pilot CGI/RPG application and best case response time is over 2
seconds (and I do not even want to discuss worst case). This is on a
server with plenty of capacity.

Well, the reason I don't like to compare application speeds is because you're rarely comparing apples to apples. Web applications in particular have lots of moving parts, and a bad design decision anywhere along the line can kill you. What I shoot for instead is the theoretical best case just to prove that it's not inherently the JSP Model 2 architecture (or in this case the EGL/JSF plumbing) that is at fault when an application runs poorly.

In my case, I'm running a persistent connection to an RPG program that doesn't set on LR. By doing this, I'm taking advantage of the incredible architecture that allowed us to run hundreds of users in a memory space measured in megabytes (or less, sometimes <grin>) with a processor that couldn't outrun a TRS-80. In fact, if you move the JVM off the machine by putting the appserver on either a Wintel or *nix box, there's almost no way a PROPERLY WRITTEN application can bottleneck a System i. You might need to go up a tier if you drastically increase the number of users, but typically that indicates that you're doing well, right?

However, it's very easy to break this design. A query over a file without the appropriate index, a program that sets on LR, an inappropriate use of ACTGRP(*NEW); all of these can kill an application, and none of them are easily detectable by the naked eye -- that is, during unit testing. That's because these kinds of problems typically only show up under load.

All that being said, though, if you're getting 2 second response time, then it's definitely time to start investigating. This is where debugging skills come into play, and also a good logging mechanism <smile>.

Joe