Online Cache Modeling for Commodity Multicore Processors ∗

Modern chip-level multiprocessors (CMPs) contain multiple pro-cessor cores sharing a common last-level cache, memory intercon-nects, and other hardware resources. Workloads running on sep-arate cores compete for these resources, often resulting in highly-variable performance. It is generally desirable to co-schedule work-loads that have minimal resource contention, in order to improve both performance and fairness. Unfortunately, commodity proces-sors expose only limited information about the state of shared re-sources such as caches to the software responsible for scheduling workloads that execute concurrently. To make informed resource-management decisions, it is important to obtain accurate measure-ments of per-workload cache occupancies and their impact on per-formance, often summarized by utility functions such as miss-ratio curves (MRCs)