Performance and scalability in SAP?

The performance of ALE business processes depends on many factors (type of business process, number of messages, activities running on the distributed systems, hardware, and so on).

To obtain detailed information about the performance a sizing procedure is required.

The sizing procedure for SAP systems linked via ALE can be divided into two independent steps. In the first step only the activities within one system must be taken into account, in the second step the additional load created by the communication between the systems has to be determined.

The total processing capacity (CPU, main memory, disk space and network bandwidth) needed for each individual SAP system can be calculated as the sum of the two steps of the sizing procedure.

The reason for choosing this simplified sizing model is the following. The resource consumption of a business transaction depends only little on the fact whether the transaction was executed by an online user, a batch job or an interface program processing an IDoc.

This is essentially because the same activities like consistency checks or business functions have to be performed in any case.

Of course the error handling for online users is different from the error handling in batch processing. On the other hand, for batch jobs or interface programs additional protocols have to be written to ensure restartability.

The first step of the sizing procedure is therefore either a user or a throughput based sizing for each SAP system. Thus the same input data is required and the same procedures and models have to be applied as in the standard sizing methodology for a single SAP system.

For the second sizing step - calculating the impact of the data transfer between the ALE linked systems - it is essential to know how many messages of a certain type will be exchanged between the different systems per time interval.

In the receiving system these messages will create additional objects or changes to existing objects, this can be taken into account by a throughput based sizing and has to be added to the numbers calculated for this SAP system in the first sizing step.

The overhead created in the sending system is in most cases less than 10% and can therefore be neglected.

This is due to the fact that normally messages are not sent immediately but bundled to reduce the communication overhead.

Measurements and experience show that the additional network bandwidth needed by ALE in a LAN scenario is much smaller than the network traffic between application and database server because of data compression; for WAN scenarios, however, the size and amount of the IDocs transferred has to be taken into account.

The same sizing considerations are valid for SAP systems connected to non-SAP systems, too.