Evaluating the Isolation Effect of Cache Partitioning on COTS Multicore Platforms
نویسندگان
چکیده
Tasks running on a Commercial Off-The-Shelf (COTS) multicore processor can suffer significant execution time variations due to inter-core interference in accessing shared hardware resources such as shared last-level cache (LLC). Page-coloring is a well-known OS technique, which can partition the LLC space among the cores, to improve isolation. In this paper, we evaluate the effectiveness of page-coloring based cache partitioning on three COTS multicore platforms. On each platform, we use two carefully designed microbenchmarks and perform a set of experiments, which generate very high interference at the shared LLC, with and without cache partitioning. We made two interesting findings: (1) Without cachepartitioning, a task can suffer up to 103X slowdown due to interference at the shared LLC. (2) More surprisingly, we found that cache partitioning does not necessarily eliminate interference in accessing the LLC, even when the concerned task only accesses its dedicated cache partition (i.e., all memory accesses are cache hits); we observe up to 14X slowdown in such a configuration. We attribute this to contention in the Miss Status Holding Registers (MSHRs) of the LLC.
منابع مشابه
A Novel Scheduling Framework Leveraging Hardware Cache Partitioning for Cache-Side-Channel Elimination in Clouds
While there exist many isolation mechanisms that are available to cloud service providers, including virtual machines, containers, etc. , the problem of side-channel increases in importance as a remaining security vulnerability – particularly in the presence of shared caches and multicore processors. In this paper we present a hardware-software mechanism that improves the isolation of cloud pro...
متن کاملOn the Design and Implementation of a Cache-aware Soft Real-time Scheduler for Multicore Platforms
JOHN MICHAEL CALANDRINO: On the Design and Implementation of a Cache-Aware Soft Real-Time Scheduler for Multicore Platforms (Under the direction of James H. Anderson) Real-time systems are those for which timing constraints must be satisfied. In this dissertation, research on multiprocessor real-time systems is extended to support multicore platforms, which contain multiple processing cores on ...
متن کاملOn the Energy Efficiency of Last-Level Cache Partitioning
Computing systems frequently have a mix of interactive, real-time applications and background computation to execute. In order to guarantee responsiveness, the interactive and background applications are often run on completely disjoint sets of resources to ensure performance isolation. These practices are expensive in terms of battery life, power and capital expenditures. In this paper, we eva...
متن کاملTowards the Integration of Theory and Practice in Multiprocessor Real-Time Scheduling
Nowadays, multicore and multiprocessor platforms are the standard computing platforms for desktop and server systems. Manufacturers of traditionally uniprocessor embedded systems are also shifting towards multicore platforms. This deeply influences the design of real-time systems, where timing constraints must be met. In the industrial world, the design of such systems largely relies on porting...
متن کاملSarkar, Abhik. Predictable Task Migration Support and Static Task Partitioning for Scalable Multicore Real-time Systems. (under the Direction of Frank Mueller.) Predictable Task Migration Support and Static Task Partitioning for Scalable Multicore Real-time Systems
SARKAR, ABHIK. Predictable Task Migration Support and Static Task Partitioning for Scalable Multicore Real-Time Systems. (Under the direction of Frank Mueller.) Multicores are becoming ubiquitous, not only in general-purpose but also embedded computing. This trend is a reflection of contemporary embedded applications posing steadily increasing demands in processing power. On such platforms, pre...
متن کامل