Redvers Batch Optimisation Service - Case Study
Barclays Life Assurance (UK)
Since 1988 Barclays Life Assurance had been growing a successful life and pensions business but as the number of clients passed one million, the overnight batch process began creeping into the on-line day. As well as broken SLA’s, the cost of this "creep" was estimated at £70 in lost business for every minute the on-line system was down. The company had completed a major re-engineering exercise a few years before to address the same problem and this had provided some temporary relief but now Barclays needed something more.
The Challenge
Redvers Consulting was asked to analyse the situation and come up with a list of suggested solutions for evaluation. The client’s users were adamant that no functionality was to be lost and that no policies were to be removed or archived from the database, no matter how obsolete. The mainframe computer on which the life and pensions application ran was administered by a separate division of Barclays and the service was billed on a CPU usage basis. This made it impossible to reduce physical I-O using store/search techniques.
Solutions
Statistics were gathered to identify the overnight critical path and the longest running programs. For each case in point the following questions were asked:
- Was the required input data available elsewhere on the system?
- Could the logic be changed to be more selective on input?
- Was all the processing incurred actually necessary in order to produce the required output?
- Could the logic be ported to a different program that used the same inputs?
- Did the output actually go anywhere?
As expected, the above questions found many positive responses which resulted in 19 independent solutions for evaluation, each of which could be applied in series or simultaneously.
Results
15 of the 19 solutions were implemented over 5 months, resulting in a 40% reduction in overall run times. One program went from a run time of two hours to three seconds.
Conclusion
Changes to the application over 12 years of evolution had repeatedly fulfilled business demands but ignored the needs of IT. The optimisation project gave us a chance to view the system from an IT standpoint and address these needs.
Any mature application, no matter what platform, language or industry, will have been repeatedly contorted from its original form in order to align itself with current business needs. The result is confused reactive systems instead of efficient proactive ones.