Abstract for Thomas John Name: Thomas John Organization: BITPIX E-mail: Go to the bio

Broadband Network Management using Transportable Agents

Management of high speed telecommunications networks is a difficult and challenging problem. At present, management of these networks is done using a wide variety of functional yet isolated systems. As broadband networks are pressed into greater utilization there is a need to integrate isolated management systems. The total broadband network is made up of different types of networks such as SONET, ATM and Frame Relay. Isolated management systems exist for managing different parts of this network. Despite many years of effort it seems unlikely that these efforts will lead to comprehensive integrated management systems in the near future. Meanwhile broadband network functions have become increasingly complex and expensive. We examine the use of transportable agents as a way of integrating disparate management systems.

We suggest an evolutionary agent-based solution which maintains the current isolated systems in their functions while supporting integrated operations. These functions are supported using an emerging control network being put in place to enable interoperability between management systems. Broadband network management consists of fault, performance and configuration management. We consider one of the target management areas, namely fault management, for illustrating the evolutionary model using transportable agents. By launching transportable agents from existing management systems, we show how root faults can be isolated in O(n) time. We also show how transportable agent movements can be minimized by utilizing emerging CMIP protocol-based agents in the network. Transportable agents also reduce network load in the control network while diagnosing faults, especially during alarm flood conditions.

We also consider the needs of other areas of network management, i.e. configuration and performance management, using transportable agents. Preliminary results show that this model can provide efficient solutions to certain problems such as rapid bandwidth allocation. Results also indicate that while transportability promotes control network efficiency, it is also necessary to maintain a system of state-oriented stationary agents to achieve this efficiency.