Allerton 2015 Paper Abstract

Close

Paper ThB1.6

Kadhe, Swanand (Texas A&M University), Soljanin, Emina (Bell Labs), Sprintson, Alex (Texas A&M University)

When Do the Availability Codes Make the Stored Data More Available?

Scheduled for presentation during the Invited Session "Distributed Storage II" (ThB1), Thursday, October 1, 2015, 12:10−12:30, Library

53rd Annual Allerton Conference on Communication, Control, and Computing, Sept 29-Oct 2, 2015, Allerton Park and Retreat Center, Monticello, IL, USA

This information is tentative and subject to change. Compiled on November 19, 2019

Keywords Queuing Theory and Analysis, Data Storage, Coding Theory

Abstract

In distributed storage systems, a special sub-class of locally repairable codes, known as the availability codes, has been proposed to enable recovery of each data block from multiple, disjoint groups of nodes of small size. Availability codes, thus not only provide good fault tolerance and its efficient maintenance, but also allow access to data by simultaneous users. In this paper, we first analyze the impact of availability parameters, i.e., the size and the number of repair groups, on the download latency by computing the mean download time using multiple repair groups and the probability that repair groups perform slower download than the systematic node. Then, we explore the case when the cumulative service capacity of the system is limited, and determine the optimal service capacity allocation across nodes such that mean download delay of a single request is minimized. We consider capacity allocation problem under three traffic splitting models, which differ in what fraction of requests are routed to the node storing the desired data block and to one or more of its repair groups. Throughout, we assume that the download requests arrive at a low rate. We find that although, in principle, higher availability should help in reducing download delays, this is not the case in our scenario. This suggests the necessity to further investigate the impact of availability on download latency by exploring different arrival, service, and content access models.

 

 

All Content © PaperCept, Inc..


This site is protected by copyright and trademark laws under US and International law.
All rights reserved. © 2002-2019 PaperCept, Inc.
Page generated 2019-11-19  13:02:16 PST  Terms of use