T. Eze, R. Anthony, C. Walshaw & A. Soper
This paper presents work towards a new architecture for trustworthy autonomic systems (different from the traditional autonomic computing architecture) that includes mechanisms and instrumentation to explicitly support run-time self-validation and trustworthiness. The state of practice does not lend itself robustly enough to support trustworthiness and system dependability. For example, despite validating system.s decisions within a logical boundary set for the system, there.s the possibility of overall erratic behaviour or inconsistency in the system. So a more thorough and holistic approach, with a higher level of check, is required to convincingly address the dependability and trustworthy concerns. Validation alone does not always guarantee trustworthiness as each individual decision could be correct (validated) but overall system may not be consistent or dependable. A new approach is required in which, validation and trustworthiness are designed in and integral at the architectural level, and not treated as add-ons as they cannot be reliably retro-fitted to systems. In this paper we analyse current state of practice in autonomic architecture and propose a different architectural approach for trustworthy autonomic systems. To demonstrate the feasibility and practicability of our approach, a case example scenario is examined. The example is a deployment of the architecture to an envisioned Autonomic Marketing System that has many dimensions of freedom and which is sensitive to a number of contextual volatility.
Keywords: trustworthy architecture; trustability; validation; autonomic marketing; autonomic system; dependability