🚀 go-pugleaf

  The Preliminary Oral Examination for the Degree of Ph.D. for


			UGUR CETINTEMEL

  will be held:		Thursday, July 29, 1999, 2:00 p.m.
			A. V. Williams Building, Room 4185

  Title of Proposal:	Decentralized Data Consistency Protocols for
			Mobile and Wide-area Environments  
  
  Abstract:		Data replication is crucial for performance,
availability, and mobility. There has been significant research on data
consistency protocols that keep replicas consistent to ensure application
correctness. Conventional data consistency protocols, however, are not
well-suited for mobile and weakly-connected environments since they
typically rely on synchronous update propagation and multi-site commit
mechanisms (e.g., 2PC). Such mechanisms, known to be expensive even in
highly-connected environments, are clearly impractical in mobile and
weakly-connected environments that are characterized by frequent
disconnections and failure-prone, unpredictable communications media. My
research focuses on developing asynchronous, decentralized protocols that
provide data consistency in these environments.

My work to date consists of two parts. The first study focused on
investigating the tension between partitioning- and replication-based
approaches for token-based commodity distribution applications in
wide-area environments. In the second study, I proposed a decentralized
replication mechanism specifically designed for mobile and
weakly-connected environments. The protocol eliminates several
restrictions of previous work by implementing an asynchronous
weighted-voting scheme via pair-wise information flow, and committing
updates locally and independently at each server. I characterized the
performance of the protocol using a detailed simulation study.

In my future research I will extend my data replication work that, being
an initial study, ignored issues such as multi-item updates and failure
tolerance. I propose to build a prototype system that implements the
protocol in order to validate previous simulation results as well as
investigate several tradeoffs in update synchronization. I further plan to
study adaptive synchronization and dynamic weight redistribution policies
to enhance protocol performance.


  Examining Committee:	Dr. Peter Keleher, Chair
			Dr. Jeff Hollingsworth, Dept. Rep.
			Dr. Michael Franklin
			Dr. Sudarshan S. Chawathe 

    EVERYONE IS CORDIALLY INVITED TO ATTEND THE PRESENTATION.