Key Technologies About Intelligent Node In Optical/Photonic Grid Network

With the development of science technologies, dynamic real parallel anddistributed applications require more complex dispersed heterogeneousresource accessing and initializing. Accordingly, underlying networkinfrastructure faces new challenge and demand. Optical grid makes hugecapacity optical network effective. Intensive data applications demandabundant network resource and dynamic & flexible service model whichadapts to optical network and consists with its current developingdirection for using optical fiber resources effectually.Optical grid as a grid application infrastructure exerts optical bandwidthand intelligent transmission to implement distributed storage, distributedprocessing, distributed transferring and so on, which is presented as thedefinition, aiming at resources collaborating and sharing optical networkresources widely for users. The key technologies research on intelligentoptical grid node addresses to uniform optical network and heterogeneousresources for building abroad grid resources environment in a dynamicscheduling way.Specifically, this thesis studies optical/optical grid architecture forsatisfying the diverse services, and makes deep and innovative researcheson node structure, controlling approach, resource allocation andscheduling algorithm Under the basis of optical grid researchingbackground, this thesis circles the optical grid architecture, and respectsto grid node key technologies for further analysis, and the maininnovations can be summarized as follows:(1) Two schemes are presented, including ASONGrid architecture onintelligent OCS and Grid over OBS (GOBS) architecture. As for Gridnetwork service doubles the management plane ability and user's ability is not limited to request, ASONgrid arhichitecture is given integratingwith grid middleware. Equally, considering OBS multi-granularitybandwidth, GOBS architecture frame is presented and provided edgenode extending structure for supporting grid service accessing to gridmiddleware, resource scheduling. For OBS resource scheduling,QoS-aware parallel scheduling algorithm is given. By GridFTP big filetransport service, GOBS model and traffic model are validated.(2) Aim at solving the application-driven optical network resourceallocation, optical resource on-demand parallel scheduling approach ispresent in this thesis, to realize grid manifold service on basis of resourceefficient utilization. Lightpath provisioning and wavelength allocationintroduce time-window multiplexing, and application request indicatesservice time and requiring bandwidth, which lightpaths are used indifferent time slots. Using Formulating analysis model and simulating forvalidating the presented, the results show that resource allocation ontime-windows can promote utilizing efficiency for avoiding blockingresulted from long time resource occupation in intensive dataapplications.(3) Scheduling Algorithm based on light-tree and DLT is presented.The presented algorithm considers solving how to control opticalresource for parallel and distributed applications to decrease system totalcost time. By building light-tree and distributing divisible load on it, theuniform scheduling approach can satisfy the requirement of parallel anddistributed services. The algorithm involves traffic model in virtue ofmulticasting light-tree and divisible load distributing theory, which issimulated for further illustration. The results show that DLT modelcombining with light-tree adapts to the resource scheduling demand ofapplications in optical grid.(4) On the basis of optical switching, design grid node physical frameand function structure around grid applications。For offering opticalresource allocation, grid node provides software resource controlling andcoordinating grid middleware. According to above designing, optical gridtest-bed is built; the innovations in the section are as followings:i). Constructing ASONGrid node infrastructure and function model; ii). Funded by National 863 Program project about optical grid, Opticalgrid Network test-bed node apparatus, optical grid middleware aredeveloped;iii). On the PGN test-bed, experiments about big file transform servicebased on GridFTP to validate optical grid feasibility and key characters:Application orienting, resource control integrating, resource schedulingparallelism, resources dispersed and sharing widely, multi-granularitybandwidth service, diverse GNS.