An Important Report from LinkedIn

This post on LinkedIn is extremely important for those trying to “Get Rid of the Mainframe”. Net Net: It’s the I/O, stupid. At least that’s the reporter’s opinion.

The distributed computing community dismisses the advantages of IBM’s I/O design. Moving many of the complexities of I/O device and data management off the mainframe and onto “Device Controllers”, peripheral processors with direct memory access. A great deal of progress has been made but this says, at least in this case, that there is a ways to go.

I’m a mainframe guy at heart. No longer much of an IBM lover but the z/System hardware is a potentially amazing platform for Linux, BSD, etc. And the existing OS facilities for BATCH are much superior to those Open Platforms. Heirloom Computing has done a great job of creating the framework for batch processing (JES2, JCL, etc.) and has seen great success. Their COBOL and PL/I compilers produce Java and run under Tomcat. And they run Watch just fine. Probably why they are probably the most successful at moving customer workloads to Open Platforms.

Is Java as fast as native executables? And if not is difference worth arguing about?

I’ve wasted too many hours with too many people and now refuse to even get into those discussions. The answer is that we’ll find out as GCC COBOL (gcobol) gets its performance kinks straightened out and the fragility addressed.

Well, we’ll find out if people will try it. I see that Debian has packages for gcobol-15. That means we should see gcobol-16 soon-ish. Unfortunately the performance work is in the gcobol-17 stream. If you want to play with the “Tip of the Spear” our CI/CD “GCC Master plus COBOL” builds can be downloaded as Debian/Ubuntu packages from our registry.