Re: AS400/i5 etc general behaviour

From: "Alan Campin"
Same problem you run into with Client Access SQL jobs. User
stops to take check a figure and when they come back and make
another call, the job has been swapped and now they have to wait
for it to get swapped back.

Well, given a memory constraint on server that COULD be true, but maybe not. It could be that the workstation CPU has a prolonged spike at 100% while Windows swaps memory. I've seen that quite a bit in Task Manager; a common bi-product of bloated software.

The delay could be a combination of a socket time-out and SQL connection dropping. Thick client applications also tend to download large result sets, leading to bandwidth bottlenecks. Bandwidth bottlenecks have gotten worse as more thick-client applications are deployed as Java applets, Adobe Flex, and MS Silverlight applications.

Web applications add another dynamic to the performance question. A Java application server or PHP may be performing workload management within their respective runtime environments that IBM i would perform for native applications. It's pretty common for PHP jobs to be performing their own caching and swapping of scripts.

It may be one thing to swap out a 2 meg RPG job, but another thing to swap out a 64 meg PHP job. I'm not sure how much memory a Websphere job uses, but I'm pretty sure that you never want it to ge swapped out - plus you're dealing with a multi-threaded environment that runs its own garbage collection within each job.

Also when dealing with Web interfaces, you not only have interprocess communication between browser and HTTP server, but also interprocess communication from HTTP server to various application runtime environments, and from there you have additional connections and interprocess communication to DB server jobs.

Unfortunately, the performance question is complicated, if you adopt distributed architectures.

Nathan.