Significant Savings at Nordea Liv
A weekly batch run, running in an extended version once a month, had little by little grown into somewhat of a monster job. The weekly runs were bad enough, lasting over 30 hours to complete, but the extended version of the application had reached 80+ hours elapsed, consuming 25 hours CPU.
Nordea accepted an offer from the Scandinavian GiAPA sales channel, SOSY A/S in Copenhagen: GiAPA should collect performance data when this job was active, and the results should be analyzed to determine if any optimization potential could be found.
The chart tells the results, which exceeded all expectations. GiAPA's automatic analysis of file accesses showed thousands of millions of reads of one certain file, where most records were read much more than a thousand times. This suggested that major savings probably were possible.
The involved developers were surprised when seeing the data - it had never been known that the file in question was read that much. But since the results of the run had been correct, nobody had ever detected the many superflouous reads, which appeared to be rather easy to remove.
Additional benefits obtained
Since the change to avoid unnecessary reads was made to a generally used subprogram, the modification had a positive effect on other applications as well. As the IT department could tell, interactive response time for two very frequently used transaction types was reduced by around 90 %.
Return on investment
It is close to impossible to calculate the exact value of savings obtained by optimization. Which value is correct for the 27 hours CPU time saved here? Difficult to answer, especially if the server was not overloaded throughout the entire run. But still more difficult, what is the business value of completing a job in one rather than three days, and of nn transactions being speeded up?
Here the bottom line was so remarkable that we at iPerformance believe that we can add Nordea Liv & Pension to our list of customers experiencing ROI just by optimizing one program.