Co-execution of multiple workloads in modern multi-core servers may create severe performance degradation and unpredictable execution behavior, impacting significantly their Quality of Service (QoS) levels. To safeguard the QoS levels of high priority workloads, current resource allocation policies are quite conservative, disallowing their co-execution with low priority ones, creating a wasteful tradeoff between QoS and aggregated system throughput. In this paper we utilise the cache monitoring and allocation facilities provided by modern processors and implement a dynamic cache partitioning scheme, where high-priority workloads are monitored and allocated the amount of shared cache that they actually need. This way, we are able to simultaneously maintain their QoS very close to the levels of full cache allocation and boost the system's throughput by allocating the surplus cache space to co-executing, low-priority applications.     «

Co-execution of multiple workloads in modern multi-core servers may create severe performance degradation and unpredictable execution behavior, impacting significantly their Quality of Service (QoS) levels. To safeguard the QoS levels of high priority workloads, current resource allocation policies are quite conservative, disallowing their co-execution with low priority ones, creating a wasteful tradeoff between QoS and aggregated system throughput. In this paper we utilise the cache monitoring...     »