Fast, On-Demand Software Deployment with Lightweight, Independent Virtual Disk Images
Bin Chen, Nong Xiao, Zhiping Cai, Zhiying Wang, Ji Wang · 2009
This paper proposes a novel software deployment approach based on lightweight, independent virtual disk images. The main goal of our approach is to provide a new software provision mode as convenient and flexible as that in the tool-toolbox model so as to reduce software management cost and to improve software reusability at a finer granularity in local or distributed virtual machine environments. With our approach, the traditional one-piece large-sized virtual disk image is split into a base virtual disk (BVD) image and multiple small-sized software virtual disk (SVD) images at a basic granularity of a single kind of software. By applying a file system metadata extraction (FSME) approach, most SVDs can be made fully independent. Only the small-sized SVDs installed with user-required software are distributed to users and are made usable/unusable by performing low-cost SVD operations at user sides, which could greatly reduce the software deployment cost. We have built a working prototype based on QEMU virtual machine monitor (VMM) and ext2 file system on Linux system. Experiments on the prototype system show that our approach effectively supports low-cost, on-demand software deployment in virtual machine environments.