Autonomic supervision of stigmergic self-organisation for distributed information retrieval

Abstract

This paper will consider how a network of information sources might be autonomously monitored to allow it to selfoptimise with respect to querying. While future networks will need to be able to self-adapt, the dynamic and autonomous nature of such networks will make the supervision process more difficult to implement in programming terms. Stigmergic linking is a lightweight and flexible way to provide some form of optimisation. If evaluation functions can also measure the success of any query, then it may be possible to monitor the performance of the self-optimisation. A supervision system could adjust the link update method until an acceptable balance between search time and quality of service is reached. Thus at least in this respect, autonomic supervision would be possible. The monitoring system might also monitor 'concept drift' and detect when it occurs. This measures typical boundaries for concepts of interest and detects when these boundaries are violated. When concept drift occurs, the system would be able to tell if this resulted from a fault or simply a change in the system use and thus be able to apply the appropriate solution.

A supervised self-organizing approach for large-scale multistage batch scheduling problems

“…But mostly, the self-organizing bottom-up process do not have the predictability of global properties. Thus a supervision solution in the self-organization system was proposed by Greer et al [11], and applied in distributed query optimization problems. With the supervision solution, concept drift from self-organization could be detected and dealt with, thereby the query performance was improved.…”