A parameter-influencer based model of a Self-healing network

Autonomic features can be built into systems by including Self-* properties. There is a dire need to find new techniques of building in Self-* features in networks for delivering important goals like maintaining uptime and QoS, since system adminstrators are crying for help as complexity of networks and their management systems continues to grow unabatedly and managing them effectively and efficiently turning out to be a veritable nightmare. In this paper we describe how a correlation based parameter-influencer model can be used to build in Self-healing features in a network. We describe various combinations of parameters and influencers which can be used to model features of varying complexity that can be added to networking systems. Statistical techniques like linear regression are used to identify positive and negative correlation amongst parameters and influencers and changes made by the Autonomic Manager(AM) or the network management systems (NMS) to bring the network to a state of stability and thus reduce downtime. Our model is programmed as daemon on the network and two case studies are described where historical data is used to tune the network parameters more accurately. The first demonstrates how uptime can be sustained and the second demonstrates how performance (and in turn QoS) can be maintained.