Sunday, May 30, 2010

The Computer That Never Dies

I've been reminded of the halcyon days a number of times in recent weeks. Meeting old colleagues and working on integration middleware components with legacy systems has taken me back to my days as a developer on mainframes.

For me, my mainframe of choice was a Bull DPS8000/DPS9000 which ran the GCOS8 operating system. Time-Sharing, JCL and EBCDIC were all terms I grew up with and I really enjoyed my time working on the platform. The code I wrote 20 years ago still exists which is reassuring and precisely the point of this article. The code was designed and written in a manner which meant it would stand the test of time.

Indeed, the code for that particular environment was so good that it required no modification in the run-up to the Year 2000 anti-climax.

In my early days, the mainframe took up almost an entire floor of an office block in the centre of Belfast. To be honest, this was at a time when my PC housed an enormous 8088 processor, with 12" long expansion cards and was painful to move about because of its sheer bulk. After a while, the 8088 was replaced with the 80286, then the 80386, then the 80486, then Pentium processors and so on. The mainframe adopted a similar evolution by getting quicker and smaller to the point where it no longer looked like a mainframe and took up little more space than other servers littered around the data centre.

Of course, during that period, most commentators expected the mainframe to finally die and be replaced by the modern array of UNIX based servers from HP, IBM and Sun. Why spend all that money on a mainframe when a couple of RS6000s could do the job just as well? Except they couldn't. Middleware did middleware very well. Mainframe did mainframe very well.

But the mainframe hasn't died. It's still alive and well in most of the big data crunching data centres and some would say that it has a new lease of life. Why? Because mainframe can now do middleware, that's why. The mainframe is no longer limited to long-running COBOL based batch processes invoked via JCL put together by Sys-Progs who look like dinosaurs. Mainframes can happily host our web servers, application servers and message buses as well our databases and batch routines.

But there is a problem! Mainframes aren't sexy and aren't attracting young talent into the arena. Sys-Progs are a dying breed. Guys who know their mainframes are in short supply. But at least we have tools to help the mainframe newbies administer their system.

Which brings me on to an exciting new initiative I've been privileged to be involved with. In conjunction with System Z and RACF guru, Alan Harrison (of Practically Secure fame), I've been working with building an extension to the IBM Tivoli Identity Manager interface to provide slick integration with zSecure Audit to provide a complete RACF management web based solution that integrates with an enterprise wide provisioning system. The result is now available from Pirean (where both Alan and I currently ply our trade).

Mainframes are suddenly sexy again and it almost feels like a home-coming for me.

No comments: