An Approach to Runtime Memory Leak Detection Oriented to Xen Virtual Computing Environment

The research on system performance and stability of virtual computing environment for related technology and application is of important theoretical and practical significance. Memory leaks, which takes place in a long-time-not-shutdown system, may have serious consequences in the practical application. Memory leak detection running in virtual computing environment is a challenging problem. To solve this problem, the phenomenon of memory leaks is classified. Based on the analysis of memory management technology using Xen hypervisor, an approach to runtime memory leak detection is proposed oriented to Xen virtual computing environment, which is also named VMLD (virtualization memory leak detection), and its design of overall framework are built and implemented. Corresponding detection rules are also put forward. VMLD has realized several modules such as maintainer (Xen internal buffer), controller, interceptor, and monitor by using defined six hyper calls and the modification of the virtual machine monitor. The experimental results show that VMLD method can detect the runtime memory leaks effectively, and has better performance.