A system and method for summarizing metric data with a plurality of levels of varying granularity. The system preferably comprises a distributed computing environment, i.e., an enterprise, which comprises a plurality of interconnected computer systems. At least one of the computer systems is an agent computer system which collects raw data relating to one or more metrics, i.e., measurements of system resources on the agent computer system. A Universal Data Repository (UDR) receives raw metric data from one or more agents. The UDR summarizes the raw data into a more compact yet meaningful form. The UDR can summarize both raw data and data that has previously been summarized one or more times, thus creating a plurality of levels of summarization. With each successive summarization, metric data become more compact, yet the data retain information and meaning. Each level of summarization is coarser in granularity and typically older than the previous level: the metric data representing a given period of time become more summarized and take up less space. So that the record of a particular process is never totally lost, process state changes are preserved throughout. The UDR preferably stores each level of summarization in a different file. When a file fills up to its configured maximum size, the oldest metric data from that file are summarized and pushed into the next coarsest file. When the coarsest file fills up, the oldest metric data from the coarsest file are deleted.