Fiserv Benchmarking on POWER9 Proves to Large Regional Banks That They Have Scalability to Spare
Performance testing alleviates concerns about mergers.
The financial services industry always has been highly competitive, leading many banks to consider how they can extend their reach. As a result, they’re looking at mergers and acquisitions to gain new customers, expand their geographic footprint, increase access to capital for lending and investment purposes, and generally improve business-operation efficiencies.
But with expansion comes some practical IT issues. Older systems may not scale or meet new processing requirements, which could hamper banks’ anticipated growth. With that in mind, Fiserv undertook performance benchmark testing to demonstrate how running its Premier bank platform on an IBM POWER9* system could improve batch processing times.
Setting the Bar
As part of performance benchmark testing, Fiserv used the Premier platform to handle batch processing requirements that were 4x the current levels of a large regional bank. Premier on IBM POWER9 also accommodated an online transaction volume at least 5x the current peak online transaction volume of that
Performance benchmarks were established using the bank’s current production database on POWER9. Performance benchmark testing was repeated after doubling and then quadrupling the current production database. For online transactions, requests per hour and average response times were measured using the workloads of more than 3 million and more than 5 million requests per hour, respectively.
To simulate real-world online traffic, a combination of four different kinds of online requests were tested: ATM, Business Online, Retail Online and internal Navigator for Premier. The intent was to make the workloads as realistic and as legitimate as possible, representing the kind of traffic today’s data centers are expected to support.
The results were dramatic. Benchmark testing on POWER9 demonstrated that even with a quadrupled database (simulating $120 billion in assets and 7.2 million accounts), batch processing was completed in slightly more than 4 hours. (See Figure 1)
The batch workload for these tests represented typical nightly processing at month’s end, making it a particularly intensive workload to test. The batch workload included a broad range of programs that support processing requirements for deposits, loans, safe deposit boxes, exception items, general ledger, relationship rewards, customer information file, custom programs and regulatory reporting.
Note that these benchmark performance results were achieved using a single database that was quadrupled to simulate a single financial institution of 7.2 million accounts, not four separate databases processed simultaneously or in parallel. Batch processing is more challenging when the accounts and processing are conducted as one institution.
The time required for processing the double database was less than twice that of the current database, and the time required for processing the quadruple database was less than twice that of the double database. These results demonstrate economies of scale for Premier on POWER9 as account volume and processing requirements grow. (See Figure 2).
In terms of online requests, the performance of Premier and commonly deployed ancillary solutions on POWER9 was impressive, accommodating a workload of more than 5 million requests per hour, greatly exceeding the online request workload of any current production deployment of Premier.
The workload represented a combination of requests through customer-facing channels (ATM and commercial and consumer online applications) and internal requests through Navigator (the Premier UI). (See Figure 3). The average response time for all tested workloads was less than 0.02 seconds (see Figure 4).
The results of the performance benchmark testing of Premier running on POWER9 demonstrated to the large regional bank—and extrapolates to mid-size banks—that this combination can support considerable future growth, whether organic or through mergers and acquisitions.
Post a Comment
Note: Comments are moderated and will not appear until approvedcomments powered by Disqus