Special Report: Network Provisioning

Legacy Networks: Since the DOE large-science projects are carried out by teams of geographically dispersed scientists, it is reasonable to expect that not all networks will be endowed with the newer provisioning and transport technologies. It would be more efficient to phase in newer methods into the production environments, and thus it is important to support the co-existence of various operations with the legacy networks (at least during the transition period). Such gradual transition to newer technologies can help the user adoption and faster integration into production applications.

Instrumentation and Diagnostic Tools: Considering that several of the required technologies are at the forefront of provisioning and transport areas, it is important to provide measurement and diagnostic tools both during the development and deployment phases. Such capabilities can potentially make the development processes more efficient, and can make it easier to diagnose the operational problems. For example, by utilizing the web100/net100 instruments as an integral part of TCP based methods, it would be possible to easily diagnose the problems and tune the protocol parameters. Similar instruments could be developed for other transport methods as well as provisioning methods.

6.3. Research Test-Beds Due to the extreme demands imposed by DOE large-science applications on networking, existing test-beds and simulation tools are inadequate to provide sufficiently detailed operating conditions such as physical layer losses over long distances, background traffic levels at tens of Gbps, or realistic switching times of optical equipment. More generally, there have been two major shortcomings in previous efforts to develop high-performance network capabilities.

First, there have been no test-beds to provide adequate operating conditions in terms of bandwidths, distances and traffic levels. Most simulators are not capable of supporting data rates of the order of Tbps, particularly over dedicated channels. Most existing test- beds do not provide tens of Gbps speeds with on-demand provisioned paths between application nodes that are separated by thousands of miles. Historically, methods based on simulations and small scale test-beds often resulted in technologies which fell short of the needs. In particular, simulations enable a detailed study of transport methods but mostly under small network configurations and IP connections. But such results are not extendable to high performance networks (particularly with dedicated channels), and furthermore they hide some of the subtle performance issues. The small scale test-beds are not able to accurately represent the physical losses typical of long haul high bandwidth channels and subtle timing effects of control and signaling channels.

Second, the adoption of research tools by the users has been highly limited due to the lack of natural transition paths. Tools developed by network researchers often require a significant amount of integration before they can be used by non-experts, and consequently are not deployed extensively in the field. This is particularly true of the protocols developed using special purpose simulators or test-beds, since they often have to be re-implemented from scratch in application and production environments. The provisioning and transport technologies described in the previous sections can only be adequately developed on powerful research networks or test-bed capable of providing real operating conditions. Thus, there is a need for a network test-bed that provides the following functionalities:

Previous    Next    Table of Content for report: Network Provisioning    Home

Network Provisioning

 

 

 

Photuris.com - Optical Data Networking