An Integrated Model for Performance Management in a Distributed System

Abstract

Real-time systems are growing in size and complexity and must often manage multiple competing tasks in environments where CPU is not the only limited shared resource. Memory, network, and other devices may also be shared and system-wide performance guarantees may require the allocation and scheduling of many diverse resources. We present our on-going work on performance management in a representative distributed real-time system—a distributed storage system with performance requirements—and discuss our integrated model for managing diverse resources to provide end-to-end performance guarantees.

Publication
OSPERT 2008