Conclusion

 The paper focused on evaluating the different approaches currently used for performance monitoring of distributed applications. Therefore a classification and analysis of available approaches has been done. Overall result was, that current techniques are not sufficient to solve today's service management problems. The paper concludes with open research questions in the area of application performance management that - in our opinion - are most important to be tackled in the forthcoming years.

Our current work focuses on the research questions raised above. Especially in the area of component based application development some promising prototypes are under development that might lead to a high degree of instrumentation automation. In the area of automated transaction correlation we already have achieved some first substantial progress which is about to be published soon.

Acknowledgment
The authors wish to thank the members of the Munich Network Management (MNM) Team for helpful discussions. The webserver of the MNM Team is located at http://wwwmnmteam.informatik.uni-muenchen.de.