Wavelength Division Multiplexing Optical Networks for Supporting Grid Computing
Display Full Text |
Download Full Text
1.18 MB PDF file
Grid computing is a computing model in which various resources, such as processing power, storage systems, data sources, or instruments, are interconnected by a communication infrastructure and accessible as a public utility for solving large scale resource intensive problems. Grid computing requires dynamic communication between distributed resources and high bandwidth quality assured survivable communication services. Optical fiber communication and networking with wavelength division multiplexing (WDM) provides a promising means to support the communication needs of grid computing, by offering huge capacity, relatively low latency, as well as dynamic control and allocation of bandwidth at various granularities. In this dissertation, we explore several important and challenging issues in optical WDM networks for supporting grid computing:
Diverse Routing for Survivable Service: Grid computing requires survivable communication services due to the huge traffic volume involved. We study the diverse routing problem in networks with a shared risk link group (SRLGs) failure. An SRLG includes a set of optical links that are affected by a single failure in the physical layer of a WDM network. The objective is to determine a pair of SRLG-disjoint paths with the minimum total cost.
Quality of Service Routing: To ensure the quality of service for grid communication, connections with bounded risk of failure and transmission delay are highly desirable. We study the multi-constrained routing problem that finds a path that guarantees the end-to-end failure risk bound in terms of total weight of SRLGs while minimizing the end-to-end transmission delay bound, or the path cost.
Scheduled Service Provisioning using Light-trails: The recently proposed light-trail concept is a promising technique in WDM optical networks for supporting the grid computing traffic given light-trails' dynamic provisioning ability. We model the grid communication demands as a scheduled traffic model, in which a set of demands is given and the setup time, teardown time, and the requested bandwidth of a demand are known in advance. We consider efficient provisioning scheduled services for supporting grid communication using static and dynamic light-trails.
Cost Effective Services with Waveband Switching and Light-trails: Waveband switching allows several wavelength channels to be batch switched and therefore reduces the number of ports needed in the switches, so that the complexity and cost of switching nodes can be significantly reduced. We study the waveband switching technology in conjunction with the light-trail architecture for supporting grid computing. This study provides us with critical insights into the designing problem of applying waveband switching technology into light-trail optical networks.
Joint Scheduling of Grid Tasks and Communication: Finally, we propose to tackle the service provisioning problem in light-trail WDM networks that jointly considers the grid task scheduling problem for supporting the grid computing. A flexible task model (FTM) is presented for modeling the grid tasks. This model is more general and flexible than the conventional task graph model considered in previous work. Simulation results show that our scheduling algorithm with FTM significantly reduces the total task completion time, compared with the results obtained using the conventional method. Note: This research is partially funded by DAGSI and DOE.
Document number: wright1221753325
Permalink: http://rave.ohiolink.edu/etdc/view?acc_num=wright1221753325
This ETD has been downloaded 315 times (through March 2013)
© 2008, all rights reserved.
This open access ETD is published by
Wright State University and OhioLINK.
Refworks
