Adabas Natural, Software AG`s database solution, delivered superb results during recent benchmarks tests run at IBM`s Independent European Laboratories in Montpellier France.
The tests were the result of a request by a consortium of Belgian banks to have Software AG`s implementation of IBM`s Parallel Sysplex architecture benchmarked during a product evaluation. Not only did the benchmark meet the group`s expectations, but also Adabas Natural`s results outstripped rival products by a factor of 40.
Rainer Herrmann, critical applications manager at Software AG, Germany, explained the background: "The consortium had recently acquired and merged two banks in Belgium, and asked the Bank of Luxembourg, a long-standing Adabas Natural user, if the database could improve their load tenfold. The bank`s IT manager replied in the affirmative, and Software AG was then requested to compare its solution to Cobol DB2 through an independent benchmarking test."
Performance improvements at the merged banks would be necessary if the new entity - which will need to support in excess of 15, 000 concurrent database users - is to compete successfully.
At the laboratory two fully configured Z-series IBM mainframes - each with 2 800 mips (million instructions per second) and 16 processors were made available. The processors are currently the most powerful mainframe offering from IBM.
"We ran the production load of the two banks against the database utilising IBM`s Parallel Sysplex Architecture with the new Adabas Clustering Services which uses data sharing," says Herrmann.
The results were overwhelming.
In the first of a two-part benchmark, Software AG achieved 157 679 database calls per second. This translates to 567 million calls per hour or 13.6 billion calls per day.
The second part involved simulating 15 000 concurrent users with IBM`s Teleprocessing Network Simulator (TPNS) tool. Sixty batch jobs were run in parallel and the result for this online test was 430 CICS transactions per second, translating to 25 731 per minute and 1.5 million per hour.
An average response time of 0.57 seconds was recorded.
"The bank was only asking for proof that their existing Adabas Natural system can handle a workload of 10 times the existing workload. We managed to boost this up to 40 times. Needless to say, the IT department was rather taken aback at the power of Adabas Clustering Services," says Herrmann.
The results of the benchmark will soon be presented to the consortium as part of a full-blown evaluation process. The highly impressed board of directors at the Bank of Luxembourg has already placed an order for Adabas Clustering Services.
Dimension Data, the solutions provider for Software AG in Southern Africa, hosted Herrmann recently in SA where he ran seminars on Adabas and provided performance training.
Share