Autonomic Distributed Data Management with Update Accesses

This paper discusses and motivates autonomic replication in a data management layer for network management systems. The data management layer simultaneously supports multiple update protocols for single logical data–item, allowing different levels of performance and consistency to be offered to different groups of users. How replication is applied is key to maximising system performance. The challenge lies in the fact that the optimal scheme depends on dynamic characteristics of the workload. Autonomically reconfiguring replication in response to the changing workload using feedback control. The goal of the paper is to demonstrate the applicability of feedback control to maintaining a given performance metric in a changing environment. The presented work is intended as an investigation and proof of concept into the feasibility of autonomically managing a facet of replication: specifically, the number of replicas. Replica placement and optimality of management decisions is currently considered out of scope and part of future work.