next up previous
Next: Development architecture Up: Runtime architecture Previous: Management agent and management

Measurement interface  

The measurement interface is the interface provided by the measurement correlator which is called by both the instrumented application and the instrumented libraries. Figure [*] shows the specification of this interface (where UTA and STA mean user transaction and subtransaction resp.). As can be seen from the figure, no explicit correlator (like necessary when using ARM API) must be passed as parameter. Instead, correlation of measurements is completely automated based on the approach described in section [*]. The following paragraphs briefly explain, how the measurement interface is expected to be used:

Figure:  Specification of Measurement Interface

In any case, measurement of a user transaction starts with a call to startUTA. If the complete user transaction is executed within a single control flow, any number of startSTA/stopSTA-pairs might follow (to measure individual subtransactions). At an arbitrary point during execution, a call to stopUTA informs the measurement correlator that the result of the transaction has been presented to the user.

If additional control flows are started during execution, the measurement correlator is informed via addControlFlow. Analogously, a control flow leaving execution of a transaction causes a call to removeControlFlow.

The most complicated case is activation of existing control flows. As an example, figure [*] shows a sequence diagram of a remote procedure call (RPC). The local application uses a communications mechanism to transparently call the remote server. The communications mechanism issues an initiatedTA-call to receive a unique identifier for the current invocation from the local measurement correlator. This identifier is then transparently transmitted to the remote system and given to the remote measurement correlator as a parameter in a startSTA-call. Then the remote server is activated. While executing, the remote server again uses startSTA/stopSTA-calls (to the remote measurement correlator) in order to provide detailed information about its further subtransactions. Before returning to the calling system, the communications mechanism issues a stopSTA-call to the remote measurement correlator to inform it about the end of the subtransaction. Using the exchanged identifier, management applications later can easily correlate the remotely executed subtransactions to the appropriate user transaction.

Figure:  Example: Sequence Diagram of a remote procedure call

next up previous
Next: Development architecture Up: Runtime architecture Previous: Management agent and management
Copyright Munich Network Management Team