An expert view on mainframe migration

Vendors offering hardware replacement systems for business-critical mainframe platforms are always crowing about the cost of application software maintenance for these mammoth systems.

Yet mainframes are still used by most high street banks to enable the processing of millions of financial transactions per day.

Mainframes are costly for several reasons

Many of the costs referred to by hardware replacement vendors are incurred as a result of requests for new features in the software, and fixes for any software bugs that develop over time.

Costs can also rise if the original system architect is no longer at the company, and if the original documentation (including the software architecture design, the actual source code, and the testing of the running application - called regression testing) associated with that software has not been retained.

Even when the documents are available, whether they are complete and readable is another issue. The extent to which firms spend money on maintaining legacy mainframe software is an issue not just for banks, but for other financial services firms, telecoms companies and government organisations.

The huge cost of running a mainframe has resulted in many companies migrating from these business-critical financial services applications onto new hardware platforms.

Mainframe migration services firm DFA president Francis Feldman now helps large firms to achieve a complete application migration onto new hardware.

Previously, Feldman was vice president at the Shared Data Center (SCD) Euronext team, which clears all daily trades for the New York Stock Exchange (NYSE). The company managed a phased two-year migration from their mainframe system and began the process in 2006.

Although the economic situation in 2006 was difficult, Feldman said that the company had the economic standing and budgetary capability to begin the migration.

The end result, a reduction in costs, was the driving factor: "We expected to see a drop of between 30 and 70 per cent," he said.

Steps involved in migration

So what steps are involved in a mainframe migration? Feldman said that it is important to understand what assets the company has, both hardware and software; that a company has the source code for every executable (application running on the system); and finally that it is aware of and in conversation with all the stakeholders for the application suite(s) it is planning to migrate.

Feldman said: “Everybody knew what their role was for this phase and we had senior management buy-in and support, meaning everybody knew they had to toe their line of responsibility."

"It also meant that stress was evenly dispersed throughout the company," said Feldman.

The next stage was for the whole team to sit down and validate a 'phased' approach to migrating the mainframe applications.

After defining which applications were to be migrated, a set of steps was created for the groups responsible for each application on the new system platform. These comprised:

1. Asset catalogue and consistency assessment
2. Workflow bridging and grouping
3. Test environment preparation
4. Application, data, and utility conversion and testing
5. Regression and initial acceptance testing. Integration and final acceptance testing
6. Operations switch over
7. Data cutover
8. Running the old and new systems in parallel
9. Running the new platform

Moving an application requires recompiling the mainframe source code on the target platform, but did the NYSE Euronext team have the up-to-date source code?
"Absolutely, but we had to go back to the developer community in order to change the code slightly and re-compile it into the American National Standards Institute (ANSI) standard," said Feldman.

Asked how much certification and regression testing NYSE Euronext had to perform, Feldman said: "I can't give a definite answer to that – it's dependent on two things: the first is the operational and development aspect, the other is what is being demanded by the business community."

A key question for Feldman is how newly migrated platforms should be set to go 'live' commercially.

In the case of NYSE Euronext, applications on the old and new platforms were run in parallel for a while, the length of time depending on several factors. One was the operational complexity of each application, of which there were seven in total.
Another was the demand of the business relying on the application.

“Ultimately, the businesses were the main driver and had the last word, and they told us that they wanted to feel comfortable with the new system the weekend after the production deployment,” said Feldman.

When it was all done, Feldman conceded that the cost savings were pretty good. “We managed savings of over 50 per cent per year. This was a clear incentive to approach [the migration] in a very serious manner," said Feldman.