Special Report: Network Provisioning

Architecture and Infrastructure Issues: Due to the highly exploratory nature of several components needed both in provisioning and transport areas, a number of new host and network architecture issues (that are not typical in Internet environments) must be investigated. To sustain ultra high data rates, OS-bypass, zero-copy, Remote Direct Memory Access (RDMA), and other non-conventional implementations for network technologies must be investigated to avoid the undue load on host processors just to support networking operations. Also, several computations and data generation operations might take place on clusters, and in some cases clusters might be used to generate aggregate streams at Tbps rates. To support such operations striping methods would be required to aggregate and separate the transport streams. It is also very important to provide ubiquitous monitoring and measurement capabilities as a part of the infrastructure to assist in diagnosis, debugging and performance optimization of various components.

3. Workshop Details The stated goal of the workshop was to: "address the research, design, development, testing and deployment aspects of transport protocols and network provisioning as well as the application- level capability needed to build operational ultra-speed networks to support emerging DOE distributed large-scale science applications over the next 10 years". It is to be emphasized that the required network capabilities must be available to the end-users most of whom are from various science areas. It is particularly desirable to minimize the demands on the users in utilizing the networking capabilities by providing suitable and transparent application interfaces. The following are the component tasks for the workshop in addressing the above goal: 1. 1. Assess the viability of existing transport protocols in supporting ultra high-speed data transfers over very long distances for distributed Petabyte datasets. 2. 2. Assess whether current core network technologies are adequate to meet the diverse network requirements of large-scale science applications. 3. 3. Assess the role of industry and federally funded network research program in developing the advanced networking technologies that meet the needs of large-scale science applications. 4. 4. Identify the major network technologies that need to be developed, enhanced or replaced in order to build and operate cost-effective networks capable of supporting distributed largescale science applications

This workshop was not intended to be a general research workshop to address a wide class of Internet problems but only to specifically address the needs of DOE large-science applications. The participants were tasked, within their areas of expertise, to: (a) identify the major bottlenecks in meeting the large-science networking capabilities, (b) identify the critical technical topics and directions, (c) outline a roadmap to develop the required networking technologies to meet the performance requirements, and (c) identify critical interfaces and connections with other areas to deliver the end-to-end solutions to application users. 3.1 Network Provisioning and Transport Areas Network provisioning and transport are two of the most critical areas representing the immediate and major bottlenecks in achieving the required networking capabilities. But at the same time they hold an enormous potential to contribute to these capabilities. A major objective of provisioning is to provide a lower layer capability to support high bandwidth on-demand and dedicated end-to-end channels. A major objective of the transport is to optimally utilize the provisioned channels to achieve stable and controlled ultra-high throughputs at the application- level. In order to tackle the above issues posed to the workshop attendees, two parallel groups were formed:

Previous    Next    Table of Content for report: Network Provisioning    Home

Network Provisioning

 

 

 

Photuris.com - Optical Data Networking