Computer Science Speaking Skills Talk
Speaker
FRANCISCO MATURANA
Ph.D. Student
Computer Science Department
Carnegie Mellon University
When
-
Where
Virtual Presentation - ET
Description
Erasure codes are typically used in large-scale distributed storage systems to provide reliability against failures. In this setting, a set of k data chunks is encoded using an [n, k] code to generate n chunks that are then stored on different nodes. Recent work shows that the failure rates of storage devices vary significantly over time, and that changing the rate of the code (via a change in the parameters n and k) in response to such variations provides significant reduction in storage overhead. However, when using traditional codes, the overhead on CPU, disk IO, and network bandwidth of realizing such a change in code rate is prohibitively high. Motivated by this application, we present a new framework to formalize the notion of code conversion, i.e., the process of converting data encoded with an [n^I, k^I] code into data encoded with an [n^F, k^F] code while maintaining desired decodability properties. As part of this framework, we also introduce convertible codes, a new class of code pairs that allow for code conversions in a resource-efficient manner.
Presented in Partial Fulfillment of the CSD Speaking Skills Requirement.