| Diving Into
An Ocean Of Storage
by David Pescovitz
store globally. That's the aim of the Berkeley computer scientists
who are pooling computing resources around the planet to create
a massively-distributed hard drive. Aptly named OceanStore, the
system could someday protect billions of users' data from earthquake,
fire, or, well, a crashed hard disk. This month, OceanStore earned
inventor John Kubiatowicz a place on Scientific American
magazine's annual list of fifty individuals and organizations whose
accomplishments demonstrate a "clear, progressive view of the technological
Engineering and Computers Sciences professor John Kubiatowicz.
"OceanStore is a utility service like electricity," says Kubiatowicz, a professor of Electrical Engineering and Computer Sciences. "You save your data on what looks like your local hard drive but it is really a large distributed service running in the background. The data is then copied and spread around the world so it's extremely hard to destroy."
Elements of OceanStore were recently deployed on PlanetLab, a shared
global testbed of 100 networked sites around the world, including
UC Berkeley. Sponsored by Intel Research, PlanetLab serves as a
digital laboratory for the development of new distributed network
services. Within six months, Kubiatowicz hopes to have OceanStore's
archival component running constantly on PlanetLab.
The durability of OceanStore data comes from encoding the bits in a data hologram, similar in principal to an optical hologram where a small piece is all that's needed to reconstruct the entire image. When a user saves a file to OceanStore, it's encrypted, duplicated and
mathematically chopped into dozens of unique fragments. Each fragment is then
sent across the Internet to geographically-dispersed OceanStore service
providers for safe keeping.
Here's an extreme example: you save a priceless digital photo of your newborn
to OceanStore. Five years later, simultaneous earthquakes, fires,
and hacker attacks around the globe knock out every copy but 16
of the 64 fragments that your photo was chopped into. Thanks to
the algorithm used to split the data, known as an "erasure code,"
those 16 fragments are enough to rebuild the entire file.
In addition to keeping your data safe, OceanStore is designed to keep your files near you. The OceanStore application runs on a novel networking layer called Tapestry. Unlike today's Internet, where each physical machine on the network has a numerical address, Tapestry generates a unique identification tag for each document. That way, a user visiting Hong Kong can access the closest copy of their data instead of, say, connecting to their regular server in Berkeley.
"Fully-assembled versions of your most active data follow you," Kubiatowicz says. "That's a performance benefit because the data can be fetched more quickly if it's in the same city and not across an ocean."
In the longterm, Kubiatowicz foresees OceanStore becoming a key ingredient in distributed sensor networks, a major thrust of the UC Berkeley-based Center for Information Technology Research in the Interest of Society (CITRIS). The CITRIS idea is that these societal-scale networks of tiny wireless sensors could monitor everything from environmental conditions to the structural integrity of buildings. But like our precious digital photos, those billions of bits of data will need a place to call home.
"OceanStore is the storage piece of the ubiquitous computing philosophy," Kubiatowicz says.
Kubiatowicz's Home Page
Lab Notes is published online by the Public Affairs Office of the UC Berkeley College of Engineering. The Lab Notes mission is to illuminate groundbreaking
research underway today at the College of Engineering that will dramatically change our lives tomorrow.
Editor, Director of Public Affairs: Teresa Moore
Writer, Researcher: David Pescovitz
Designer: Robyn Altman
Subscribe or send comments to the Engineering Public Affairs Office: firstname.lastname@example.org.
© 2002 UC Regents.