Performance optimization remains one of the key issues in parallel computing. With the emergence of large clustered SMP systems, the task of analyzing and tuning scientific applications actually becomes harder. Tools need to be extended to cover both distributed and shared-memory styles of performance analysis and to handle the massive amount of information generated by applications on today’s powerful systems. This paper proposes a flexible way to define hierarchies of event streams and to enable the end-user to traverse these hierarchies, looking at sampled or aggregated information on the higher levels. The concept will be implemented and evaluated in practice within the scope of the US DOE ASCI project.