On the timeliness of a cluster based large scale online video surveillance

Koushik Sinha, Atish Datta Chowdhury, Subhas K. Ghosh, Satyajit Banerjee

Research output: Chapter in Book/Report/Conference proceedingConference contribution

2 Citations (Scopus)

Abstract

Timeliness is an important issue for video based surveillance and is often quantified by the delay between the time of availability of image frames from cameras and completion of their processing. Most existing commercial video surveillance systems focus on the issues of efficient storage and retrieval, remote monitoring, data streaming, forensics and limited real-time analysis - but not explicitly on the timeliness issues of large scale online analysis vis-a-vis resource utilization. In this paper we present a new load distribution strategy for on-line, large scale video data processing clusters that are used as an aid to manual surveillance. We propose a novel approach for fine grained load balancing, modeled as a minimization of average completion time problem. The proposed approach is robust in the sense that it is not dependent on the estimates of future loads or on the worst case execution requirements of the video processing load. Simulation results with real-life video surveillance data establish that for a desired timeliness in processing the data, our approach reduces the number of compute nodes by more than a factor of two, compared to systems without the load migration heuristics.

Original languageEnglish
Title of host publicationProceedings of the IASTED International Conference on Parallel and Distributed Computing and Systems
Pages47-52
Number of pages6
Publication statusPublished - 1 Dec 2008
Externally publishedYes
Event20th IASTED International Conference on Parallel and Distributed Computing and Systems, PDCS 2008 - Orlando, FL, United States
Duration: 16 Nov 200818 Nov 2008

Other

Other20th IASTED International Conference on Parallel and Distributed Computing and Systems, PDCS 2008
CountryUnited States
CityOrlando, FL
Period16/11/0818/11/08

Fingerprint

Processing
Resource allocation
Cameras
Availability
Monitoring

Keywords

  • Average completion time
  • Cluster computing
  • Load balancing
  • Online video processing
  • Parallel system
  • Resource utilization
  • Total completion time
  • Video surveillance

ASJC Scopus subject areas

  • Software
  • Hardware and Architecture
  • Computer Networks and Communications

Cite this

Sinha, K., Chowdhury, A. D., Ghosh, S. K., & Banerjee, S. (2008). On the timeliness of a cluster based large scale online video surveillance. In Proceedings of the IASTED International Conference on Parallel and Distributed Computing and Systems (pp. 47-52)

On the timeliness of a cluster based large scale online video surveillance. / Sinha, Koushik; Chowdhury, Atish Datta; Ghosh, Subhas K.; Banerjee, Satyajit.

Proceedings of the IASTED International Conference on Parallel and Distributed Computing and Systems. 2008. p. 47-52.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Sinha, K, Chowdhury, AD, Ghosh, SK & Banerjee, S 2008, On the timeliness of a cluster based large scale online video surveillance. in Proceedings of the IASTED International Conference on Parallel and Distributed Computing and Systems. pp. 47-52, 20th IASTED International Conference on Parallel and Distributed Computing and Systems, PDCS 2008, Orlando, FL, United States, 16/11/08.
Sinha K, Chowdhury AD, Ghosh SK, Banerjee S. On the timeliness of a cluster based large scale online video surveillance. In Proceedings of the IASTED International Conference on Parallel and Distributed Computing and Systems. 2008. p. 47-52
Sinha, Koushik ; Chowdhury, Atish Datta ; Ghosh, Subhas K. ; Banerjee, Satyajit. / On the timeliness of a cluster based large scale online video surveillance. Proceedings of the IASTED International Conference on Parallel and Distributed Computing and Systems. 2008. pp. 47-52
@inproceedings{f2f8c12553424b2d986adabb8300a43d,
title = "On the timeliness of a cluster based large scale online video surveillance",
abstract = "Timeliness is an important issue for video based surveillance and is often quantified by the delay between the time of availability of image frames from cameras and completion of their processing. Most existing commercial video surveillance systems focus on the issues of efficient storage and retrieval, remote monitoring, data streaming, forensics and limited real-time analysis - but not explicitly on the timeliness issues of large scale online analysis vis-a-vis resource utilization. In this paper we present a new load distribution strategy for on-line, large scale video data processing clusters that are used as an aid to manual surveillance. We propose a novel approach for fine grained load balancing, modeled as a minimization of average completion time problem. The proposed approach is robust in the sense that it is not dependent on the estimates of future loads or on the worst case execution requirements of the video processing load. Simulation results with real-life video surveillance data establish that for a desired timeliness in processing the data, our approach reduces the number of compute nodes by more than a factor of two, compared to systems without the load migration heuristics.",
keywords = "Average completion time, Cluster computing, Load balancing, Online video processing, Parallel system, Resource utilization, Total completion time, Video surveillance",
author = "Koushik Sinha and Chowdhury, {Atish Datta} and Ghosh, {Subhas K.} and Satyajit Banerjee",
year = "2008",
month = "12",
day = "1",
language = "English",
isbn = "9780889867741",
pages = "47--52",
booktitle = "Proceedings of the IASTED International Conference on Parallel and Distributed Computing and Systems",

}

TY - GEN

T1 - On the timeliness of a cluster based large scale online video surveillance

AU - Sinha, Koushik

AU - Chowdhury, Atish Datta

AU - Ghosh, Subhas K.

AU - Banerjee, Satyajit

PY - 2008/12/1

Y1 - 2008/12/1

N2 - Timeliness is an important issue for video based surveillance and is often quantified by the delay between the time of availability of image frames from cameras and completion of their processing. Most existing commercial video surveillance systems focus on the issues of efficient storage and retrieval, remote monitoring, data streaming, forensics and limited real-time analysis - but not explicitly on the timeliness issues of large scale online analysis vis-a-vis resource utilization. In this paper we present a new load distribution strategy for on-line, large scale video data processing clusters that are used as an aid to manual surveillance. We propose a novel approach for fine grained load balancing, modeled as a minimization of average completion time problem. The proposed approach is robust in the sense that it is not dependent on the estimates of future loads or on the worst case execution requirements of the video processing load. Simulation results with real-life video surveillance data establish that for a desired timeliness in processing the data, our approach reduces the number of compute nodes by more than a factor of two, compared to systems without the load migration heuristics.

AB - Timeliness is an important issue for video based surveillance and is often quantified by the delay between the time of availability of image frames from cameras and completion of their processing. Most existing commercial video surveillance systems focus on the issues of efficient storage and retrieval, remote monitoring, data streaming, forensics and limited real-time analysis - but not explicitly on the timeliness issues of large scale online analysis vis-a-vis resource utilization. In this paper we present a new load distribution strategy for on-line, large scale video data processing clusters that are used as an aid to manual surveillance. We propose a novel approach for fine grained load balancing, modeled as a minimization of average completion time problem. The proposed approach is robust in the sense that it is not dependent on the estimates of future loads or on the worst case execution requirements of the video processing load. Simulation results with real-life video surveillance data establish that for a desired timeliness in processing the data, our approach reduces the number of compute nodes by more than a factor of two, compared to systems without the load migration heuristics.

KW - Average completion time

KW - Cluster computing

KW - Load balancing

KW - Online video processing

KW - Parallel system

KW - Resource utilization

KW - Total completion time

KW - Video surveillance

UR - http://www.scopus.com/inward/record.url?scp=74549135137&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=74549135137&partnerID=8YFLogxK

M3 - Conference contribution

SN - 9780889867741

SP - 47

EP - 52

BT - Proceedings of the IASTED International Conference on Parallel and Distributed Computing and Systems

ER -