- Overview: What's a Mix?
- Real-World Low-Level Technology Stack Test Input and Mixes
- Testing and Tuning for Daily System Loads
- Testing and Tuning for Business Peaks
- Identifying Key Transactions and Business Processes
- Real-World Access Method Limitations
- Best Practices for Assembling Test Packages
- SAP Component and Other Cross-Application Test Mix Challenges
- Tools and Approaches
9.8 SAP Component and Other Cross-Application Test Mix Challenges
For each particular SAP application, component, or solution, there tends to be a special set of circumstances or limiting factors that complicates creating a repeatable, consistent, or adequate test mix. The following list identifies key problem areas I've run into in the past, and what I did to work around these issues:
R/3 business processes range from one-offs to multicomponent, highly complex sets of transactions. But, for most companies, R/3 activities tend to focus on discrete transactions run repeatedly by many users throughout the day, alongside a subset of core batch processes. I suggest focusing on a few key business processes, like order-to-cash, to take advantage of the input/output nature of the underlying transactions. In this way, obtaining input data is a simple matter of leveraging the previous transaction's output. And the first transactiona sales ordercan originate in any number of other systems and be easily completed if core customer, material, and plant data are available and abundant.
APO includes both online users and batch-oriented business processes. The real load borne by the system is represented by the latter, though, especially with regard to demand or production planning runs, or Available to Promise (ATP) processing. Online user activity is negligible in comparison. From an input data perspective, the number of key figures that reside in your liveCache server, the number of characteristic combinations (properties) that describe an object, the number of periods (measured in weeks) against which these processes will run, and of course the number of sales, purchase, planned, and other orders transferred from R/3 to APO are all important.
BW and SEM activities range tremendously, so it's no surprise that input data ranges are equally wide. In the past, I've started a BW stress test by kicking off an R/3 data extraction process, or simply by pulling data from the ODS to create and populate a cube. These are certainly important functions, but may not represent your most important goals. Instead, you may wish to run queries against one or more standard or custom InfoCubes. The key will therefore be the cubes themselvesif it matters little how the data move through your SAP system landscape, make it easy on yourself and start with a fully populated set of cubes.
CRM data hail from a number of systems within the CRM system landscape, including the TRex Server, Multi-Channel Interface Server, InQMy Application Server, and potentially a Workgroup Server and Communication Station, not to mention the CRM server itself. As such, it can be one of the more complex SAP solutions to stress test. As a starting point, I suggest that you concentrate on the core user type your system hosts (probably a Mobile Sales or CIC user, though others exist), and determine the input necessary to support the top five or so key user transactions. For your Mobile Sales users, you might focus on managing opportunities or activities, creating customer orders, performing service-related transactions, or managing customer, product, or project-specific data, for instance. If you host CIC users, you might instead go with the transactions and activities outlined in the SAP CRM benchmark kit, for example, and focus on transactions relevant to managing incoming and outgoing CIC calls
EBP (the core component of SRM) is quite complex, especially from an input perspectivedata originating from Requisite BugsEye (an online catalog), R/3, BW, and EP complicate executing core EBP business processes. But that's not all! There are other SRM components that need to be considered as well, like the SAP Bidding Engine and SAP Supplier Self Services. Given all of this, a seemingly straightforward shopping-cartdriven procurement business process will have many touch points, will pull data from many sources, and will generally represent a whole lot of scripting and coordination effort.
EP has evolved significantly over the last few years and today represents one of the fastest growing products in SAP's line-up. It's also potentially one of the most important products a company will implement, in that all users could conceivably (and, it is hoped, will) leverage its single sign on capabilities to gain access to all other SAP systems and many third-party resources, creating a potential high-availability and performance nightmare should the EP technology stack lack scalability. Thus, stress testing EP will only grow in importance as it continues to be deployed across the globe. Fortunately, stress testing can be conducted quite easily. Sure, many systems represent potential integration points, but an effective load test could very well consist of accessing only local resources.
PLM supports users responsible for managing product, asset, and process information at any point in the product lifecycle, from selection and purchasing through production ramp-up, installation, operation, engineering changes, maintenance/repair, retirement, and more. From a data input perspective, then, you need to understand the precise functionality being implemented and what then needs to be testedeverything from Life-cycle Data Management and Asset Lifecycle Management to core functionality like Program and Project Management, Quality Management, and Environment, Health, and Safety (or EHS) can come into play. Plus, because PLM is implemented as an enterprise portal solution and is tied closely with CRM and likely APO, the business processes that can result may be complex indeed. I therefore suggest going after either the biggest couple of functional areas to be implemented (in terms of transaction counts or user counts), or instead the most critical functional areas, and script core transactions rather than full-blown business processes.
The world of SAP has grown considerably over the last year, thoughwell beyond the components and solutions just highlighted. The SAP XI, xApps, and other products represent additional systems that may be tested individually or as part of a larger SAP solution. For example, simulating the number and size of the messages that the XI Integration Server processes represents an excellent method of testing XI without all of the integration points necessarily in the picture.
XI offers some exciting capabilities when it comes to viewing collaborative cross-application SAP business processes. For example, business processes that span your R/3, APO, and CRM systems can be tied together via synchronous and asynchronous messaging defined by and maintained within XI. You may then leverage XI to enable true cross-application business process stress testing, regardless of whether heterogeneous system landscapes have been deployedtest cases would initiate from one of the core SAP systems tied together in this manner, of course. And test execution would be seamless, because XI would handle moving the output from one transaction into the proper component, where it becomes input for the next transaction in a business process. Beyond this awesome timesaving by-product of XI integration, XI also allows you to visually depict and manage SAP components as well as other enterprise applications from Baan, Broadvision, JDE World Software, Oracle, PeopleSoft, Siebel, and more. In this way, the inclusion of SAP and third-party enterprise applications may all be tied together into a cohesive virtual system. More to the point, these complex systems may be managed as a single entity rather than as a bunch of individual systems, which further simplifies testing.
Mainstays like SAP's ITS also continue to thrive. From a data input perspective, I suggest you focus on common transactions supported by back-end systems like R/3. Because all (or nearly all, in the case of older releases of R/3) SAP screen content is maintained in HTML on the ITS server, testing ITS amounts to testing the speed and throughput of Web connections. Fortunately, tools abound that support this input type, as we reviewed in Chapter 6, and scripts of this nature are fairly easy to write and maintain. And the best news of all is that you'll find that your ITS users tend to execute the same transactions as their SAPGUI-enabled colleagues (though exceptions existI've got a customer that has pushed the bulk of its HR and ESS workload on ITS, for instance, whereas other modules and functional areas are accessed via the traditional SAPGUI). So once you've established the core transactions executed by a particular SAP product or component, you're well on your way to using this information in support of ITS testing, too.