Аннотация:In this work we propose a design for a new distributed modular monitoring system framework, which allows combining both monitoring tasks (supercomputer ‘health’ and performance monitoring) in one monitoring system. Our approach allows different part of monitoring system to process only the data needed for the task assigned to this part. Another feature of our framework is the ability to calculate performance metrics on-the-fly, dynamically creating processing modules for every job or other objects of interest.