Keynote talk III

III- Atomic Shared Objects for Distributed Systems: Consistency, Latency, Reconfiguration

Alexander A. Schwarzmann
University of Connecticut

Biography:
Alexander A. Schwarzmann earned his B.S. from Stevens Institute of Technology in 1979, M.S. from Cornell University in 1981, and Ph.D. from Brown University in 1992, all in Computer Science, and he did his post-doctoral work at MIT from 1995 to 1997. His research is in fault-tolerant distributed computing and security of electronic election systems. Prior to pursuing his academic career he worked at Bell Labs and Digital Equipment Corp. From 1997 he is at the University of Connecticut, where he is now serving the Department Head of Computer Science & Engineering. He chaired and served on the Program Committees of more than 50 leading conferences, and he served as the Steering Committee Chair of both the ACM Symposium on Principles of Distributed Computing (PODC 2012-2015) and EATCS Symposium on Distributed Computing (DISC 2004-2007). Since 2006 he is also the Director of the UConn Center for Voting Technology Research (VoTeR). The Center provides technological expertise in security and integrity of electronic election systems. Schwarzmann is an Associate Editor of Information & Computation. He is an author of over 150 technical publication and three books.

Abstract:
Consistent shareable data services supporting atomic (linearizable) objects provide convenient building blocks for distributedsystems. In general it is notoriously challenging to combine provable correctness guarantees with efficiency in networked systems subject to delays and processor crashes. To deal with crashes one must replicate objects at multiple network locations, and this creates the challenge of guaranteeing consistency. We survey work on specification and implementation of consistent read/write data objects and algorithms, focusing on fault-tolerance and latency. Then we describe a framework for dynamic consistent data services that can be tailored to yield implementations for various target network settings and that incorporates on-the-fly reconfiguration that only modestly interferes with on-going read and write operations. Here the goal is to guarantee safety (atomicity) for arbitrary patterns of asynchrony, crashes, and message loss, while enabling practical implementations. We describe examples of specification, reasoning about correctness, provable optimizations, and implementations of consistent data services in distributed systems.

Dates

November 26, 2017

Abstract submission deadline

December 03, 2017

Paper submission deadline

January 20, 2018

Acceptance notification

February 24, 2018

Camera ready copy due

Proceedings

SpringerLNCS

The proceedings of the conference will be published in Springer’s Lecture Notes in Computer Science

Partners & Sponsors (Pending)

SponsorsNETYS28