Codes, Graphs, and Hyperplanes in Data Access Service
Emina Soljanin (Rutgers University)
Abstract: Distributed computing systems strive to maximize the number of concurrent data access requests they can support with fixed resources. Replicating data objects according to their relative popularity and access volume helps achieve this goal. However, these quantities are often unpredictable. Erasure-coding has emerged as an efficient and robust form of redundant storage. In erasure-coded models, data objects are elements of a finite field, and each node in the system stores one or more linear combinations of data objects. This talk asks 1) which data access rates an erasure-coded system can support and 2) which codes can support a specified region of access rates. We will address these questions by casting them into some known and some new combinatorial optimization problems on graphs. We will explain connections with batch codes. This talk will also describe how, instead of a combinatorial, one can adopt a geometric approach to the problem.
combinatorics
Audience: researchers in the topic
Carleton Combinatorics Meeting 2021
Series comments: See the external webpage for more details: people.math.carleton.ca/~dthomson/CCM2021
To contact, please use CCM2021@math.carleton.ca and prefix your subject with "[CCM2021-request]"
| Organizers: | Daniel Panario, David Thomson*, Qiang Wang |
| *contact for this listing |