Events Calendar

CS Colloquium: Srivatsan Ravi (ISI USC) - Synchronization using Transactions: Lower bounds and Algorithms

Speaker: Srivatsan Ravi, ISI USC

Talk Title: Synchronization using Transactions: Lower bounds and Algorithms

Series: CS Colloquium

Abstract: Designing algorithms to exploit today's distributed computing platforms ranging
from general-purpose multicore CPUs, cloud infrastructures to domain-specific decentralized computing systems emphasizes the need for designing robust and fault-tolerant synchronization protocols. However, traditional techniques for synchronization are either too coarse-grained to exploit concurrency or require application-specific fine-grained synchronization.

The Transactional Memory (TM) abstraction is a synchronization mechanism
for multicore programming proposed as a middle ground: it intends to combine an easy-to-use programming model with an efficient utilization of hardware concurrency. TM allows the programmer to speculatively execute sequences of shared-memory operations as atomic in-memory transactions with safe semantics: state witnessed by each transaction is consistent with some sequential execution. Thus, the programmer can design applications having largely sequential semantics in mind and let TM take care, at run-time, of dealing with problems associated with process asynchrony and adversarial failures.

In this talk, we focus on a model for hybrid TMs that exploits hardware extensions in prevalent CPU architectures to support small transactions. We present lower bound proof constructions for implementing safe hybrid transactions and its implications for the complexity of concurrent data structures. We conclude by outlining how transactions as a synchronization mechanism can serve as highly robust universal constructions for domain-specific distributed computing models.

This lecture satisfies requirements for CSCI 591: Research Colloquium. Please note, due to limited capacity, seats will be first come first serve.

Biography: Srivatsan Ravi is a computer scientist at the Information Sciences Institute in University of Southern California (USC). His research interests are centered around the theory and practice of distributed computing. Specifically, he works on algorithms and lower bounds for fault-tolerant distributed systems. His research is motivated by emerging new hardware trends that require a new abstract computation model or via introduction of distributed computing techniques to domains where the sequential implementation continues to be state-of-the-art.

He received his Ph.D. degree from Technical University of Berlin in Germany, where he received the Marie Curie Ph.D. Fellowship and was a member of Deutsche-Telekom Labs, Berlin. His Masters degree is from Cornell University, U.S.A and his Bachelors degree is from Anna University, India.


Host: John Heidemann

Location: Ronald Tutor Hall of Engineering (RTH) - 115

Audiences: Everyone Is Invited

Contact: Assistant to CS chair