The Revenge of the Mainframe.
IN THE OLD DAYS Dinosaurs ruled the world.
To some people, Mainframes appeared to be slow and unresponsive, unsuitable for the fast pace of modern life. So organisations invested in Midrange systems like Unix, and NT.
But these days, Mainframe Hardware is back - more lean-and-mean than ever. Sun "Serengeti" , HP "SuperDome", and Compaq "Wildfire" are examples of large powerful systems which often have "partitioning" capability. A large system can have multiple "virtual machines" residing on it. Each virtual machine controls it’s own memory, CPU, disks etc, yet shelters in the safety of a Mainframe-style Chassis.
IBM (of course!) has got into the act with the SP2 Frame. Looking like a rack of Mid-Range Systems, it is in fact a collection of closely coupled "nodes", able to be merged, partitioned and re-configured from one central console. IBM is now taking this a stage further, with the "Federation Switch", which brings Mainframe-style control to every IBM system in the Data-Centre, whether part of an SP2 or not.
Microsoft is also heralding the Mainframe’s return, by supporting Compaq, Unisys and others with "Windows 2000 DataCenter". Built like a mainframe, these systems are a far cry from the NT systems we knew. The control procedures are extremely rigid :- no applications go onto a W2K Datacenter system without passing rigorous testing procedures. Using this approach, suppliers are prepared to bet money on 99.999% uptime.
So is it just Mainframe Hardware that is coming back ? Not at all. Not just Hardware, but also some of the Practices of Mainframe days are returning, in a new guise.
Practice - makes Perfect ?
YEARS AGO, Organisations threw out their Mainframes and rushed into Midrange and Client-Server solutions. "Take control away from the IT Department and give it to the Business", was the cry. So the old 'Frames began to die off.
But perhaps we forgot something in our haste to become a more flexible, responsive organisation? There was much to commend the IT Departments of the past.
For one thing, systems management was easier. Instead of hundreds of small systems (all configured differently, all subject to upgrades, change, re-configuration, monitoring etc.), why not have centralised control of the Data-Centre? Why not implement proper Capacity Planning, so we know where the resources are being used, and by which application ?
No one is suggesting that we return to the "bad old days". But perhaps we could learn from them. Some of the good things about mainframes (centralised control, good management practices etc.) are making a welcome return. Sure it was slow to implement changes in the past. But how about a few procedures to ensure that we have proper testing, and Release Control? We may have thrown off the shackles of bureaucratic slavery, but did we end up with uncontrolled anarchy instead?
The Microsoft "Windows 2000 DataCenter" product serves to highlight some of the potential of a Mainframe-type approach. The control procedures are tight, with thorough testing and control of releases. Every application has to be checked to determine how it handles memory, for example. As for upgrading the device drivers from your favourite Internet site :-- forget it!
So has the wheel turned full circle? Not exactly. The Mainframe is returning. Not the old Dinosaur. But the good habits which we threw away, in an unguarded moment.
Dennis Adams, June 2001