Time Bands for Dependable Systems
time scales, social time, psychology of time, predictability
Complex computer-based systems are required to function at many different timescales (from microseconds or less to hours or more). Time is clearly a crucial notion in the specification (or behavioural description) of computer-based systems, but it is usually represented, in modelling schemes for example, as a single flat physical phenomenon. Such an abstraction fails to support the structural properties of the system, forces different temporal notions on to the same basic description, and fails to support the separation of concerns that the different timescales of the system facilitate. Just as the functional properties of a system can be modelled at different levels of abstraction or detail, so too should its temporal properties be representable in different, but provably consistent, timescales.
Time is both a means of describing properties of structures and is a structuring mechanism in its own right. To make better use of `time', with the aim of producing more dependable computer-based systems, it is desirable to identify identify a number of distinct `time bands' in which the system is situated. Such a framework enables the temporal properties of existing systems to be described and the requirement for new or modified systems to be specified. The concept of time band comes from the work of Newell in his attempts to describe human cognition. Newell focusses on hierarchical structures within the brain and notes that different timescales are relevant to the different layers of his hierarchy. By contrast, in the DIRC work, we put the notion of a time band at the centre of any description of system structure. It can then be used within any organisational scheme or architectural form - for they all lead to systems that exhibit a wide variety of dynamic behaviours.
In all bands, a common set of temporal phenomena and patterns of behaviour are likely to be exhibited by the system itself or its environment. For example, periodic (or regular or cyclic) activities, event handling (responding to an event by a deadline), temporal reasoning (planning and scheduling), interleaving and multi-tasking (and other aspects of concurrency), pausing (or delaying), analysis of response (or completion) time, deadline driven activities, and various aspect of dynamic behaviour such as rates of change. Whilst evident in all bands, these phenomena are not identified using the same terminology in the various time bands of interest (i.e in the technical, psychological and sociological literature).
In this work we have provide an informal description of a framework built upon the notion of time bands. We consider the properties of bands, how activities within a band can be organised and evaluated, and how relationships between bands can be captured and described. Note that the number of bands required and their actual granularity is system-specific; but the relationships between bands, we contend, exhibit important invariant properties. We have also begun to develop a formal model for time bands and have applied the concepts to an existing socio-technical system.
A.Burns, I.J.Hayes, G.Baxter and C.J.Fidge, Modelling Temporal Behaviour in Complex Socio-Technical Systems, University of York Report, YCS 390, 2005.
A. Burns and G. Baxter, Time Bands for Systems Structure,
for Dependability: Computer-Based Systems from an Interdisciplinary
Perspective", (editors) D. Besnard, C . Gacek and C. B.
Alan Burns (York)
|Page Maintainer: email@example.com||Credits||Project Members only||Last Modified: 30 May, 2006|