Basics of Workload Analysis and Runtime Analysis

Statistics such as response times, memory use, and database accesses are collected and stored for all transaction steps by every SAP NetWeaver Application Server (AS). The data collected for a request is called a single statistics record.

Data in the statistics record

In principle, one or more statistics records are written for every request. The content of these records can be differentiated by key figures and characteristic parameters. Key figures are measurement values-for instance, specific run-times, transferred data volumes, and the number of calls of specific actions in a program. They specifically indicate the time a component required for processing and the time the component waited for the subsequent request to be processed. Information about the point of time, user, instance name of the component, computer, type of service, and program executed is saved as characterizing parameters. 

Use cases

We differentiate two different use cases for statistics records:

->The workload analysis, which is explained in this chapter and with which you evaluate the performance data of your system periodically or via event-driven management. For this purpose, the single records are organized in load profiles and can be analyzed with the workload monitor. The workload monitor enables you to obtain a comprehensive overview of load distribution between and within SAP components. 

->The run-time analysis of individual requests . Here, you create an event-driven run-time analysis of specific requests: the statistics records of the calls to be analyzed are compiled in an analysis transaction according to their call sequence. This single record analysis enables you to determine the time elapsed in individual components and operations and which detail trace needs to be switched on next or analyzed if it has already run simultaneously .

ABAP server statistics

The statistics records of the ABAP server are written by the server's kernel. In principle, a record is written for every request processed by the server. In addition to the main record, which is always generated, the ABAP server writes sub-records for specific actions. Examples are sub-records for database calls or database procedure calls, which are intensively used, for instance, in SAP LiveCache; remote function calls (RFCs); and HTTP calls.

For performance reasons, the statistics records are initially stored in the main memory and then written to the file system on the hard drive. In a background process that runs every 60 minutes (SAP performance collector), the data is organized in load profiles based on characterizing parameters; these can be time profiles, server profiles, or transaction profiles. The single records are deleted after a specific period of time. The aggregated data is saved in the database.

Workload monitor

You can use the workload monitor (Transaction ST03) to evaluate the aggregated data. To display the single records stored in the file system, use Transaction STAD.  If you use SAP Solution Manager, it reads the data from the connected systems via a remote function call and stores it in the Business Intelligence system of SAP Solution Manager. There, the evaluation reports are preconfigured, and you can use them directly after installation.

Programming models and components

The statistics main record and its sub-records cover the following programming models and components

->Normal" ABAP transactions and reports
->ABAP transactions with SAP GUI controls
->ABAP web services such as Business Server Pages, Web Dynpro ABAP, and the internal Internet Transaction Server (ITS) 
->SAP NetWeaver BW reporting 
->RFCs between SAP systems 
->SAP liveCache
->SAP Virtual Machine Container (VMC) 

The creation of statistics and load profiles is always activated on the ABAP server to enable monitoring of the system. The process is critical for performance and has constantly been improved by SAP, so the creation and processing of statistics has grown even in SAP systems with very high throughput. SAP license agreements even specify that these statistics are required for ensuring support.