Timing QoS and Fault Tolerance Issues in ATM Networks

Students: Bin Wang, Hung-Ying Tyan, and Yi Ye

Faculty: Chao-Ju Jennifer Hou

Sponsors: NSF CAREER Program

There has been an increasing need of timely and dependable communication services either for embedded real-time applications (such as air-traffic control, automated factories and industrial process controls) or for interactive distributed services (such as multimedia conferencing and video/audio virtual realities). Performance objectives used in conventional networks --- such as maximizing the throughput or minimizing the response time --- are not of the most important concern for both types of applications. Instead, predictable timing quality of service (QoS) performance must be guaranteed. Moreover, such performance guarantees must be ensured even in the case of network node/link failure. The goal of the proposed research is thus to develop and demonstrate an environment -- an integrated set of techniques, protocol suites, and software layers -- for providing deterministic timing guarantee and high service availability on an end-to-end basis for high-speed point-to-point networks.

Central to the design and implementation of communication services with high service availability and timing QoS guarantee is the notion of virtual paths/virtual circuits (VPs/VCs). The use of VPs/VCs leads to several potentially attractive features, e.g., smaller total processing requirements, faster processing per circuit, better use of network resources, and fast path restoration capability. However, these features can not be realized without (1) careful layout of VP connections, (2) judicious selection of routes for VCs with QoS requirements, and (3) detailed design of a fast failure restoration mechanism that takes advantage of the VP/VC concept. Therefore, we consider in this project the following innovative research tasks:

Unified framework for timeliness and high availability

We develop in an analytic framework the following components:

• a fast restoration mechanism for equipping each primary VP with a backup VP and for quickly redirecting packets/cells from a failed primary VP to the corresponding backup VP upon node/link failure.
• a route selection scheme for establishing VCs with real-time QoS requirements through a variable number of logical links, where a logical link is either a VP or a physical link.

Software implementation and experimentation

To empirically evaluate the effectiveness of the proposed scheme/mechanism, we are implementing them as software layers on a POSIX-compliant operating system. Such an implementation will facilitate us to export a rich, well-defined API through which the proposed communication services are provided.

Laboratory testbed development and technology demonstration

We have built at the Ohio State University a laboratory testbed which consists of a Myrinet LAN of nine SUN workstations interconnected by 4-port and 8-port Myrinet switches and Myrinet/Sbus host interfaces. This laboratory testbed is used for development of network software and for technology demonstration on a representative embedded real-time or interactive multimedia application.

Return to Project Home Page
Date last modified -- September 30, 1996
Direct comments concerning this WWW site to: jhou@ece.osu.edu