Special Report: Network Provisioning
Ultra High-Speed transport Protocols and Services Group: The objective is to develop recommendations for protocols that can deliver and sustain multi-Gbps throughputs to the scientific applications. This working group was chaired by Wu Feng and Don Towsley. It consisted of fifteen members, 7 from national laboratories, 6 from universities, and 2 from industry.
3.2 Workshop Organization This was a "working" workshop with focused discussions on very specific problems, methods, and potential solutions in the transport and provisioning areas. The workshop started with very short introductory presentations that identified the needs and the problem space. The rest of the workshop then consisted of meetings in two parallel tracks until the last joint session. The participants provided a balance of expertise from universities, industry and national laboratories in representing the needs, technologies, research areas and business aspects. It was recognized that providing high-performance networking capabilities to the large-science application users requires the development of new and novel technologies. Furthermore, these technologies must be tested and deployed in production infrastructures. Hence it was essential that the participants as a whole represent a broad spectrum of research, academic and industrial viewpoints. There were altogether 32 participants.
Ten from national laboratories: Oak Ridge National Laboratory (3), Argonne National Laboratory (2), Los Alamos National Laboratory (2), Pacific Northwest National Laboratory (1), Stanford Linear Accelerator Center (1), and ESnet (1). Eleven from universities: University of Massachusetts (1) , Georgia Institute of Technology (2), University of Virginia (1), University of Illinois at Chicago (2), Indiana University (1), University of Tennessee (1), University of California at Davis (1), Pittsburgh Supercomputer Center (1), and California Institute of Technology (1). Eight from Industry: Celion (1), Cienna (1), Cisco (1), Juniper (1), Level3 (1), Lightsand (1), MCNC (1), Qwest (1).
Three from DOE Headquarters (two via access grid). The provisioning working group consisted of 14 participants and the transport working group consisted of15 participants.
4. Workshop Findings in Dynamic Provisioning Area Network provisioning generically refers to various aspects of a class of lower layer services of the (conventional) protocol stack to support applications. When dedicated channels are provided to the applications, however, the conventional view of the network protocol hierarchy is blurred, and additional middleware and/or transport modules are needed by the applications to utilize the provisioned channels. In general, the modes for the provisioned channels can range from the dark fiber at the lowest level through photonic switching, DWDM, Synchronous Optical Network (SONET), Generalized Multiprotocol Label Switching (GMPLS), to IP at the highest level. Depending on the mode of operation, the precise meaning of the provisioning varies; for example, it can stand for either a SONET link setup on-demand with a specified rate or a connection over conventional IP links with specified bandwidths. Based on the type of channel between two hosts, the provisioned path may be composed of different types of components such as switches, routers, service provisioning platforms, line cards, and Network Interface Cards (NICs). Typically, for IP networks such paths consist of Ethernet cards at the hosts connected to local hubs or switches which in turn are connected to router blades; the routers themselves can be connected through line cards to each other in various ways, for example via SONET or GigE links. Note that since channels can be provisioned under various modes, appropriate higher level mechanisms must be utilized to suitably expose and provide their functionalities to the transport, middleware and applications modules.
Previous Next Table
of Content for report: Network Provisioning Home
