Fast, inexpensive content-addressed storage in foundation

  • Authors:

  • Sean Rhea;Russ Cox;Alex Pesterev

  • Affiliations:

  • Meraki, Inc.;MIT CSAIL;MIT CSAIL

  • Venue:
  • ATC'08 USENIX 2008 Annual Technical Conference on Annual Technical Conference
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Foundation is a preservation system for users' personal, digital artifacts. Foundation preserves all of a user's data and its dependencies--fonts, programs, plugins, kernel, and configuration state--by archiving nightly snapshots of the user's entire hard disk. Users can browse through these images to view old data or recover accidentally deleted files. To access data that a user's current environment can no longer interpret, Foundation boots the disk image in which that data resides under an emulator, allowing the user to view and modify the data with the same programs with which the user originally accessed it. This paper describes Foundation's archival storage layer, which uses content-addressed storage (CAS) to retain nightly snapshots of users' disks indefinitely. Current state-of-the-art CAS systems, such as Venti [34], require multiple high-speed disks or other expensive hardware to achieve high performance. Foundation's archival storage layer, in contrast, matches the storage efficiency of Venti using only a single USB hard drive. Foundation archives disk snapshots at an average throughput of 21 MB/s and restores them at an average of 14 MB/s, more than an order of magnitude improvement over Venti running on the same hardware. Unlike Venti, Foundation does not rely on the assumption that SHA-1 is collision-free.