Previous Contents Next

Chapter 10   Conclusion

Correlating alarms is a task that requires good working knowledge of the system in question. In building a program to take care of monitoring a network a prime consideration should be how to make the system easy to update. This is especially important in the rapidly changing environment of a telecommunications network.

Getting this knowledge into a computer program is one part of the programming process. This thesis has discussed ways to start a knowledge base from a model of the network. There is work on maintaining a knowledge base through repeatedly generating new knowledge bases for each change. This way is a little drastic and not a satisfying solution to the problem.

The way to maintain a knowledge base, chosen for the implementation, was Ripple-Down Rules. This serves to place each rule in the context for which it was written. The structure may be limiting, but the benefits are great. Maintenance can be carried out on a daily basis by the network operator. This is a great improvement over conventional production systems which require a programmer to implement modifications plus extensive testing to make sure the system still works.

10.1   Contributions made

To my knowledge, Ripple-Down Rules has not been applied to network fault management. The closest would be the help desk for the unix system [Compton, 2000]. But a help desk is not the same as a network monitor. Compton himself, the great authority on RDR, has not heard of a case like it.

As shown in this paper, the existing methods of maintaining a rules base are inadequate. The very idea of stand alone rules controlled by consistency checking and conflict resolution is difficult to make accurate and make easy to update. Ripple-Down Rules seems to hold the answer.

Another branch of my work was to venture into the world of Machine Learning. The problem with starting an RDR knowledge base from scratch is, well, that you have to start from scratch. Machine Learning helps to bridge the gap between having some concepts and having the start of a good rules base.


Previous Contents Next