Evaluation of Distributed Control and Signaling Infrastructure for ATM Networks
Project Award Date: 0000-00-00
Two of the most important problems which must be solved before large-scale, multi-service ATM networks can become a practical reality are the efficient provision and management of services, especially effective available bit rate (ABR) services, and adaquate support for network management computations. The focus of this project is on the creation of a testbed within which the performance of network services and network management can be evaluated. This has required the creation and integration of a number of tools, some developed under other projects, to enable us to consider all factors affecting network performance.
The tools which are part of the testbed include NetSpec, the ATM Reference Traffic System (ARTS), Call Generation(CG), and the Data Stream Kernel Interface(DSKI). NetSpec is a tool for describing and conducting network level performance evaluation experiments. ARTS is a system designed to record and produce packet level ATM traffic reproducibly. The CG is designed to set up and tear down ATM connections according to various load patterns. The DSKI is a device Driver which provides a general interface for collecting event traces form the operating system. Integrated into a single suite under the control of NetSpec, we can now describe and conduct performance evaluation experiments addressing the role of every level of the user, operating system, and network signaling software in determining the performance of an ATM connection.
The architecture of next generation management software (NMS) will be required to support an increasingly complex set of services, and must be able to continuously evolve. The NMS will be object oriented, and subject to significant real-time performance constraints. The TINA standards are addressing these issues, but they are still very general and so not consider implementation details which will have a vital influence on network performance. Our work is addressing this by creating the ability to test the perfor mance of CORBA based applications under NetSpec. This will enable us to consider how each component of the total system affects the performance of NMS structures described by TINA, and thus what type of computational resources will be required to man age next generation networks using TINA based control architectures.
Primary Sponsor(s): Sprint