Coding for Communications
C.9.4
Lecture
Streaming Codes

Online Versus Offline Rate in Streaming Codes for Variable-Size Messages

Michael Harrison Rudow, K. V. Rashmi

Date & Time

01:00 am – 01:00 am

Abstract

Providing high quality-of-service for live communication is a pervasive challenge which is plagued by packet losses during transmission. Streaming codes are a class of erasure codes specifically designed for such low-latency streaming communication settings. We consider the recently proposed setting of streaming codes under variable-size messages which reflects the requirements of applications such as live video streaming. In practice, streaming codes often need to operate in an ``online'' setting where the sizes of the future messages are unknown. Yet, previously studied upper bounds on the rate apply to ``offline'' coding schemes with access to all (including future) message sizes. In this paper, we evaluate whether the optimal offline rate is a feasible goal for online streaming codes when communicating over a burst-only packet loss channel. We identify two broad parameter regimes where, perhaps surprisingly, online streaming codes can, in fact, match the optimal offline rate. For both of these settings, we present rate-optimal online code constructions. For all remaining parameter settings, we establish that it is impossible for online schemes to attain the optimal offline rate.


Presenters

Michael Harrison Rudow

Carnegie Mellon University

K. V. Rashmi

Carnegie Mellon University

Session Chair

Viveck Cadambe

Penn State University