Online set packing and competitive scheduling of multi-part tasks

Yuval Emek*, Magnús M. Halldórsson, Yishay Mansour, Boaz Patt-Shamir, Jaikumar Radhakrishnan, Dror Rawitz

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

We consider a scenario where large data frames are broken into a few packets and transmitted over the network. Our focus is on a bottleneck router: the model assumes that in each time step, a set of packets (a burst) arrives, from which only one packet can be served, and all other packets are lost. A data frame is considered useful only if none of its constituent packets is lost, and otherwise it is worthless. We abstract the problem as a new type of online set packing, present a randomized distributed algorithm and a matching lower bound on the competitive ratio for any randomized online algorithm. Our bounds are expressed in terms of the maximal burst size and the maximal number of packets per frame. We also present refined bounds that depend on the uniformity of these parameters.

Original languageEnglish
Title of host publicationPODC'10 - Proceedings of the 2010 ACM Symposium on Principles of Distributed Computing
PublisherAssociation for Computing Machinery (ACM)
Pages440-449
Number of pages10
ISBN (Print)9781605588889
DOIs
StatePublished - 2010
Event29th ACM SIGACT-SIGOPS Symposium on Principles of Distributed Computing, PODC 2010 - Zurich, Switzerland
Duration: 25 Jul 201028 Jul 2010

Publication series

NameProceedings of the Annual ACM Symposium on Principles of Distributed Computing

Conference

Conference29th ACM SIGACT-SIGOPS Symposium on Principles of Distributed Computing, PODC 2010
Country/TerritorySwitzerland
CityZurich
Period25/07/1028/07/10

Keywords

  • Competitive analysis
  • Multi-packet frames
  • Online set packing
  • Packet fragmentation

Fingerprint

Dive into the research topics of 'Online set packing and competitive scheduling of multi-part tasks'. Together they form a unique fingerprint.

Cite this