More Transparency Needed For Open Source Running on IBM i
July 25, 2018 Alex Woodie
Open source may be the future of IBM i. It certainly seems that way at the moment. But if open source is going to soar to new heights on the platform, it will need better integration with the existing processes in place to monitor and manage the platform.
That’s the opinion of JK Grafe, the CEO of Quad Nova Group, a Jacksonville, Florida-based IBM i consultancy that has offices up the Eastern Seaboard. By Grafe’s own admission, open source is a great thing for the platform. But the difficulty in seeing what’s actually going on with open source workloads is a problem.
“I can open up a PASE environment to a couple of Java programmers, and they become invisible,” Grafe tells IT Jungle. “If you have a system operator who does his WRKACTJOBs, he doesn’t see these people. He doesn’t know what they’re doing. They’re invisible to him.”
That lack of transparency can lead to performance issues down the road, especially if you open up the IBM i server to people who don’t have deep experience with running workloads on the system.
“I’m not saying it happens automatically, but it can happen,” he continues. “If they use Java and SQL and not knowing exactly how the system works and having low experience on the system, you can get in danger with performance.”
The German native is a believer in open source, even going so far as to say that most new development on the platform should be done in open source. He’s a fan of Node.JS and how it leverages the parallel nature of the IBM Power processor to handle big workloads, especially compared to Java’s tendency to spawn new JVMs for each new process.
“Just to be crystal clear — this is where we’re going. The future is open source on that box,” Grafe says. “But customers should be careful and use experienced people. If you transition from legacy to open source, there are a lot of bumps you can run into.”
Grafe has been working with the IBM i for decades – long before it was called the IBM i. He has made a good business out of developing applications for clients, including in traditional environments like RPG and newer environments like Node.JS. So to some extent, Grafe is simply marketing his company’s services.
But he has a bigger point, which is that IBM could do more to shed light on how open source software – all of which runs within the PASE AIX environment – runs on the IBM i server.
“It would be good if IBM could have some feature enhancements in their default WRKACTJOB tools and stuff like this,” he says. “Bigger shops, like hospitals and factories, have default setups. They have system operators who have a very specific skillset how to monitor a system. And if there’s a Node.js service running . . . a Node.js service on a System i just appears as a single job. There’s really no tool, so far as I know, to really see what’s going on underneath in that service. It would be nice to have some extra options.”
Likewise, the IBM i ISVs who develop management and monitoring tools don’t have a good way to get at the open source workloads that are increasingly cropping up on the IBM i server. “There are definitely some APIs that IBM needs to provide if ISVs want to provide tools for it,” Grafe says. “I think IBM has to come up with a couple of APIs for us, otherwise we can’t dig that deep under the hood ourselves.”
Grafe has been getting value out the PASE environment for many years. One of his favorite is a Linux tool that converts PDFs to TIFFs and “probably is not supposed to run on i.” However, it makes sense to use it. “It stops us from re-inventing the wheel, which RPG programmers have done so many times.
But venturing into the open source world is not a good idea for the uninitiated, Grafe says.
“It’s a little bit new territory for a lot of people. I have done PASE for a lot of time and I have combined systems in both worlds. I know the dangers,” he says. “If you want to have a safe environment and a reliable system, and not have people scratching their heads when performance goes down, there should be complex guidance on how to do these things.”
RELATED STORIES
We’ve got `ps` on PASE, we need `top` and `lsof` … PASE can provide the info you need if your sysprog is capable of interpreting it … but not all of the tools seem yet to be ported.