The Performance Improvement of an Enhanced CPU Scheduler Using Improved D_EDF Scheduling Algorithm

Abstract

With the development of cloud computing, mobile device and industry automation, virtualization plays an important role today. The core of virtualization is hypervisor which directly determines the performance of platform. Therefore, how to allocate resource effectively becomes an important issue. Xen, one of widely used open source projects, is a virtual machine monitor. Simple EDF (Earliest Deadline First) scheduler that is a dynamicpriority real-time scheduler in Xen implements the famous EDF scheduling algorithm. Due to EDF scheduling strategy has miss the deadline and inefficiency in overloaded condition, we have improved the D_EDF scheduling algorithm that combines Deadline-Monotonic scheduling strategy with EDF scheduling in virtualized environment. And we have extended the Simple EDF scheduler which provided by Xen using improved D_EDF scheduling algorithm. Our experiment demonstrates that the proposed scheduling algorithm increased the performance under CPU-intensive and memory-intensive workload in overloaded condition.