Optimizing SAP Performance

Feature Article | July 6, 2010 by Ralph K. Treitz

800 Flugzeuge, 500 Kunden: Bei SR-Technics muss SAP ERP reibungslos funktionieren. 8Foto: SR Technics)

800 aircrafts at SR Technics: Business processes have to run smoothly (photo: SR Technics)

Poor performance, stability, and availability can have various causes in SAP systems. In some industries, the “only” culprit is data quantities that are growing at a disproportionate rate. Large retail chains, for instance, process millions of movement data records in their SAP applications every day.

In most cases, however, the reasons behind drops in performance are more complex. The Swiss aviation services provider SR Technics, for example – which shares real-time information on the 800 aircraft it supports with around 500 customers – counts on its central SAP ERP application to always be up and running at a high level of efficiency. “Our business processes have to run smoothly; we can’t risk any reduction in performance,” explains Adrian Wirth, vice president of SR Technics’ IT group. Due to its constantly increasing amounts of data, the company planned and carried out a migration of its SAP application from an aging UNIX platform to a modern, highly accessible hardware platform with a corresponding storage solution. However, the move did not produce the performance boost SR Technics had hoped for. Quite the opposite, in fact: Despite the platform’s more powerful components, the SAP software’s dialog response times and batch program runtimes were even longer than before. “The performance improvements we were expecting from the migration to more potent hardware never materialized, so we had to switch back to our old infrastructure a few days later,” Wirth recalls. Then the investigation into the cause of these events began.

Hardware not always at fault

As this example shows, performance shortfalls in SAP applications cannot always be traced back to a lack of hardware resources. Individual issues can combine to cause greater problems that place unnecessary strain on these systems. In the case described, incorrectly configured hard drive systems, unstable clusters, and performance problems related to the ZFS file system were responsible.
On their own, each of these problems would have been compensated by SR Technics’ new and improved hardware; only the combination of all three led to the company’s serious performance issues. The nature of this conundrum was also such that even more powerful hardware would not have made the slightest difference.

Next Page: VMS Benchmarks

VMS Benchmarks

Why do IT managers often find it hard to pinpoint the actual causes of performance deficits? They often suspect a single reason, make a corresponding change, and typically dismiss the reason out of hand when no improvement occurs. Meanwhile, the attempted modification may actually be part of the solution. The concurrence of multiple causes often amplifies their effects, but since this amplification is nonlinear, it can be difficult or impossible to gauge. This exacerbates the problem.

Assessing individual system reports is not enough to get to the root of the issue; software that performs holistic analysis of SAP systems and their architecture is required. Model-generated monitoring based on benchmarks makes this possible (see text box).

Just such an assessment gave SR Technics key insights into why its SAP ERP system had not exhibited the performance expected following its hardware upgrade. The results recommended modifying the company’s hardware configuration and system architecture, as well as adjusting these aspects to users’ actual behavior. “Since implementing the VMS recommendations, we’ve come to expect fast response times, rapid batch processing, and high uptime in all of our IT processes. Our SAP users are also pleased,” Wirth says in summary.

Targeting complexity

One of the most important conclusions to draw from this case is that there is no universal means of optimizing SAP system performance. The demands business-critical processes place on SAP systems in terms of processing speed, response times, availability, and stability vary by situation. Companies in the automotive industry, for instance, have to be able to handle forecast and sequenced delivery schedules for just-in-time and just-in-sequence processes in seconds. For telecommunications providers, fast response times at call centers are a competitive advantage. At the beginning of each new month, publicly traded companies are obligated to submit their reports for the previous month – the generation of which involves a series of batch processes.

All of these requirements are different; even more importantly, they almost always come in groups. This leads to interdependencies, which makes it very difficult to apply the solution for a particular issue to other cases. Specific problems constantly manifest themselves in different areas.

SR Technics has demonstrated what to expect: When combined, application complexity, the abundance of contingencies among individual components, and limited resources augment one another, leading to nonlinear performance. Since this complexity cannot be avoided – indeed, it will actually increase due to virtualization and other modern technologies – companies must learn to deal with it. In complex systems, improvements only take hold when executed jointly in a coordinated manner. A single action will not have the desired effect, and the series of measures needed depends on the case at hand. Only a complete system analysis capable of factoring in as many aspects as possible can deliver a comprehensive overview and identify the elements needed to achieve optimization.

Tags: ,

Leave a Reply