January 13, 2006

MS Research Tech Report - BitVault: a Highly Reliable Distributed Data Retention Platform

MSR-TR-2005-179
BitVault: a Highly Reliable Distributed Data Retention Platform
by Zheng Zhang; Qiao Lian; Shiding Lin; Wei Chen; Yu Chen; Chao Jin
December 2005
http://research.microsoft.com/research/pubs/view.aspx?tr_id=1035

Abstract:

In this paper, we report the design and implementation of the storage layer of BitVault: a content-addressable retention platform for large volume of reference data - seldom-changing information that needs to be retained for a long period of time. BitVault uses "smart brick" as the building block to lower the hardware cost. However, the challenges are to maintain low management cost in a system that needs to scale all the way from one brick to tens of thousands of bricks, to ensure reliability and to deliver with a simple enough design. Our design incorporates P2P technologies for its self-managing and self-healing capabilities and uses massively parallel repair to reduce vulnerability window of data loss. The simplicity of the architecture relies on an eventu-ally reliable membership service provided by a perfect one-hop DHT (distributed hash table), and its object-driven repair model yields last-copy recall guarantee: independent of how many other failures that may occur and their sequences, as long as the last copy of a data object still remains in the system, the data can be retrieved and its replication degree fully restored. A proto-type has been implemented. Theoretical analysis, simulations and experiments are conducted to validate the design of BitVault.
Entry categories: Microsoft
Posted by Jorgen Thelin at January 13, 2006 01:00 PM - [PermaLink]