Survey on a combined approach using prediction and compression to improve pre-copy for efficient live memory migration on Xen

Abstract

Cloud computing is an emerging technology of cluster systems based on virtualization, which makes hypervisor to interact directly with the hardware. Live migration techniques of Virtual Machine (VM) can be useful for load balancing, energy efficiency and fault tolerance. Xen hypervisor is widely used in virtualization for its better performance and integration with different architectures. In this paper, the pre-copy approach of Xen is evaluated for Live Migration of virtual machines. The Framework of Xen-4.1 based pre-copy is configured using DRBD (Distributed Replicated Block Device) on Intel-VT architecture to get a practical understanding of the algorithm. Xen's pre-copy can have better results in terms of performance metrics-downtime, total transfer time, total pages transferred if bitmap structure and writable working set are improved. For avoiding repeated retransmission of memory pages, techniques based on prediction are used to get information about dirty pages. Prediction based two different techniques are discussed: i) LRU Stack Distance ii) Probability Prediction. Memory Compression, a faster migration technique, is the second step of a combined approach. Various compression methods named RLE, Huffman Coding, MEMCOM, WKdm and LZ are classified. Prediction and compression based combined approach is surveyed to improve pre-copy for efficient live memory migration on Xen.