Check out the new USENIX Web site.

Experiences with Content Addressable Storage and Virtual Disks

Anthony Liguori IBM Linux Technology Center

Eric Van Hensbergen IBM Research


Efficiently managing storage is important for virtualized computing environments. Its importance is magnified by developments such as cloud computing which consolidate many thousands of virtual machines (and their associated storage). The nature of this storage is such that there is a large amount of duplication between otherwise discreet virtual machines. Building upon previous work in content addressable storage, we have built a prototype for consolidating virtual disk images using a service-oriented file system. It provides a hierarchical organization, manages historical snapshots of drive images, and takes steps to optimize encoding based on partition type and file system. In this paper we present our experiences with building this prototype and using it to store a variety of drive images for QEMU and the Linux Kernel Virtual Machine (KVM).

Eric Van Hensbegren 2008-11-04