Micro Focus' fundamental business started with providing a Windows development environment for mainframe COBOL/CICs developers (CICS is IBM's mainframe Customer Information Control System for transaction processing) – why waste expensive Mainframe mips (millions of instructions per second) on developing code? That gave Micro Focus useful multiplatform COBOL development tools – and its customers then started asking why they couldn't offload the mainframe by running production work in its PC development environment.
Well, some work was needed to make this industrial-strength but now Micro Focus has a Lift and Shift strategy (http://www.microfocus.com/solutions/liftandshift/), for moving workload from the mainframe to Wintel and other environments, and it presented this to existing and potential customers at a half-day seminar in London recently.
Micro Focus recommends a two-phase approach – move applications to a more cost-effective platform first and prove that it all works properly – and only then add function and extend the application. The migration itself may prove more difficult than you expect – Andy Sinclair (senior director product management at Micro Focus) says there can be many “call and forget” assembler routines in legacy software you only find when you begin to “lift and shift” – even when the company concerned is convinced it has no assembler left.
Then there's the receiving platform. Horst Katter (Senior Product Marketing manager, enterprise Solutions EMEA at Intel) told the seminar about new features of Intel's 64bit Itanium chips that allowed them to support mainframe-style processing more effectively. Windows itself needs 64 bit addressing in order to address more cache, but mainframe applications may also need the virtualisation capabilities and “Active Management Technology” features (which let you maintain a server when it is down) of the new Itanium architecture - areas where legacy Intel architecture is somewhat broken, although (as Katter admits) “mainframes have been doing this for 20 years”.
The new processor's multicore architecture also addresses Wintel's environmental issues. Instead of ever-increasing processor speeds with associated increasing power consumption you can increase processing power and throughput at lower processor speeds by putting more cores in the CPU – and you also get better processor utilisation (mainframes happily manage over 95 per cent utilisation while older Wintel processors typically run at about 10-25 per cent). In other words, “Lift and Shift” is becoming more feasible because new processors are catching up with the capabilities of traditional mainframe architectures.
Let's get one thing straight, however – this isn't yet another “Death of the Mainframe” story. The mainframe is still a very powerful tool and mainframe applications still embody hundreds of man-years of business knowledge (learning COBOL takes a month – really – but learning banking takes years). No, this is a story about choice: about leaving some applications running on the mainframe, perhaps encapsulated as a web service; perhaps porting some to a Wintel server; perhaps replacing others with packages or even re-writing them; all driven by a careful cost/benefit analysis in business terms.
Take a typical example of a successful mainframe application. Between 1982 and 1984, the ICS (International Cooperative Solutions) CompetanceCenter of DirectGroup Bertelsmann developed an application to serve the needs of some 3.5 million book club customers in Australia, Switzerland, Italy, French Canada and Poland – about a million lines of COBOL, using a hierarchical database and representing around 1600 man-days of effort. According to the CIO of ICS CompetenceCentre, Günter Bodner, features such as special offers, returns and language variations make book club systems exceedingly complex, yet this mainframe application coped with expansion into Eastern Europe in the early 1990s and conversion to relational database technology (for added flexibility) in the late nineties – processing some 100 million account transactions per year.
Nevertheless, approaching its second decade, it became apparent that the underlying mainframe was nearing the end of its supported life and that the business environment was changing to require graphical web-based systems. At the same time, cost reduction was an imperative and ICS couldn't afford to discard the intellectual property investment in its legacy application. It had already been using Micro Focus products in development and now the idea of running the production system on Windows in the same way proved both feasible and cost effective.
ICS has migrated its book clubs to Windows in stages since 2002 and encountered few problems – online transactions were faster on the Wintel platform and although some batch transactions were slower, overall there was a 10 per cent improvement in batch processing. The migration project paid for itself out of the savings from not upgrading the mainframe and the new systems are not only cheaper initially but cheaper to operate and maintain (ICS claims concrete savings of €50,000 a month).
"Only a few years ago we would never have contemplated this project but the Micro Focus products, which manage the COBOL and CICS run-time platforms, have matured and become very efficient," says Bodner. "This combined with the enormous progress Windows has made over the last few years provided us with the viable alternative platform we have been looking for."
So, is there a downside to all this? Well, you have to be careful with the cost analyses, as ICS was. It's “obvious” that the operating cost of a mainframe will be more than that of a Wintel solution, but how much of that mainframe cost is due to licensing strategies rather than the technology itself? Mainframe software licenses are expensive while you're locked-in – but they might suddenly become cheaper if you have a viable migration path to Wintel. Buying another mainframe if you're running out of capacity (or your mainframe is no longer supported) is expensive – but IBM has schemes for shipping spare capacity you don't have to pay for until you use it. And a Windows server looks flexible – but how many are only running one application at a time and how flexible is a mainframe running many Linux virtual machines in logical partitions?
Similarly, the entry cost to Wintel is much less than the entry cost to a mainframe. But once you've migrated, is it possible that you are just as locked in to Wintel as you were to the mainframe, and what happens to the licence charges then? True, there's more competition on Wintel but you might find Microsoft's reported decision to make its Software Assurance licensing scheme compulsory for customers of its next generation Vista Windows operating system just a bit worrying.
Not that any of this worries Micro Focus, because it has software you'll need regardless of whether you're trying to link to a huge mainframe application as a Web Service; or porting the last tiny application from an almost redundant mainframe to Windows (and cancelling your mainframe license, which probably makes migration a no-brainer). Choice makes customers powerful, nowhere truer than for mainframe customers, and Micro Focus gives more choices to the customers of the mainframe vendors.