Solving a Video-Server Load Re-Balancing Problem by Mixed Integer Programming and Hybrid Variable Neighborhood Search
A Video-on-Demand system usually consists of a large number of independent video servers. In order to utilize network resources as efficiently as possible the overall network load should be balanced among the available servers. We consider a problem formulation based on an estimation of the expected number of requests per movie during the period of highest user interest. Apart from load balancing our formulation also deals with the minimization of reorganization costs associated with a newly obtained solution. We present two approaches to solve this problem: an exact formulation as a mixed-integer linear program (MIP) and a metaheuristic hybrid based on variable neighborhood search (VNS). Among others the VNS features two special large neighborhood structures searched using the MIP approach and by efficiently calculating cyclic exchanges, respectively. While the MIP approach alone is only able to obtain good solutions for instances involving few servers, the hybrid VNS performs well especially also on larger instances.
KeywordsMixed Integer Linear Program Neighborhood Structure Variable Neighborhood Search Video Server Variable Neighborhood Descent
Unable to display preview. Download preview PDF.
- 6.Walla, J.: Exakte und heuristische Optimierungsmethoden zur Lösung von Video Server Load Re-Balancing. Master’s thesis, Vienna University of Technology, Vienna, Austria (2009)Google Scholar
- 7.Yu, H., Zheng, D., Zhao, B.Y., Zheng, W.: Understanding user behavior in large-scale video-on-demand systems. In: Proceedings of the 1st ACM SIGOPS/EuroSys European Conference on Computer Systems 2006 (EuroSys 2006), pp. 333–344. ACM, New York (2006)Google Scholar
- 12.Aggarwal, G., Motwani, R., Zhu, A.: The load rebalancing problem. In: Proceedings of the fifteenth annual ACM symposium on Parallel algorithms and architectures, pp. 258–265. ACM, New York (2003)Google Scholar
- 15.Thompson, P., Orlin, J.: The theory of cyclic transfers. Operations Research Center Working Papers. Massachusetts Institute of Technology (1989)Google Scholar
- 16.Ahuja, R., Orlin, J., Sharma, D.: New Neighborhood Search Structures for the Capacitated Minimum Spanning Tree Problem. Sloan School of Management, Massachusetts Institute of Technology (1998)Google Scholar