Learn about new research in shared data and the two intriguing systems within it from Professor Jennifer L. Welch of Texas A&M University.
Invited Talk: Message-Passing Implementations of Shared Data Structures
Wednesday, October 24, 2018
Brickyard (BYENG) 210, Tempe campus [map]
Distributed storage, or shared data, is a vital mechanism for communication among processes in distributed systems and facilitates the development of higher-level applications. Although shared data is convenient in theory, it is rarely provided in large-scale distributed systems. Instead, processes keep individual copies of the data and communicate by sending messages to keep the copies consistent.
This talk covers background on the topic as well as two intriguing aspects: systems that experience churn, where the set of participating processes changes dynamically; and systems that target data structures whose specifications are relaxed.
Next, Welch discusses a recent algorithm for implementing a shared register in a dynamic system with ongoing churn works in an asynchronous system and tolerates process crashes within boundaries.
Finally, learn about distributed implementations of shared data structures with relaxed specifications. Strongly consistent implementations of shared objects with strict semantics are provably expensive, fueling interest in relaxations. A data type relaxation adds a small amount of non-determinism to the specification which can reduce the required frequency of expensive synchronization. The researchers’ results show that the algorithms are asymptotically optimal and that there is an inherent complexity gap between different levels of relaxation.
About the speaker
Jennifer L. Welch is Regents Professor and Chevron Professor II in the Department of Computer Science and Engineering at Texas A&M University. She received her Master’s of Science and doctorate from the Massachusetts Institute of Technology and her Bachelors from the University of Texas at Austin. Her research interests include algorithms and lower bounds for distributed computing systems, especially distributed shared objects and dynamic networks.