Thursday, September 25, 2003
Noon - 1 pm
Wean Hall 8220
Carnegie Mellon University
Efficient, Flexible Consistency for Highly Fault-tolerant Storage
Fault-tolerant storage systems spread data redundantly across a set of
storage-nodes in an effort to preserve and provide access to data despite
failures. One difficulty created by this architecture is the need for
a consistent view, across storage-nodes, of the most recent update. Such
consistency is made difficult by concurrent updates, partial updates made
by clients that fail, and failures of storage-nodes.
This talk will describe the design and implementation of a family of strong
consistency protocols for survivable, decentralized data storage. These
protocols exploit storage-node versioning to efficiently achieve strong
consistency semantics, and erasure-coded data for network and storage
efficiency. The protocol family is made scalable by offloading work from
the storage-nodes to the clients. The protocol family is flexible in that
it covers a broad range of fault and timing assumptions, up to asynchrony
and Byzantine faults of both storage-nodes and clients, with no changes
to the client-server interface, server implementation, or system structure.
Each protocol scales with its requirements -- it only does work necessitated
by the system and fault models.
Garth Goodson is a 6th year Ph.D. student in the Parallel Data Lab who
has been working on consistency protocols for decentralized storage for
the past year and a half. He hopes to be able to complete his Ph.D. within
the next year. Before storage protocols he worked on the Self-Securing
Storage System (S4) project, and has built numerous RPC packages.
Further Seminar Info: