Ponder2: A Policy System for Autonomous Pervasive Environments

  • Authors:
  • Kevin Twidle;Naranker Dulay;Emil Lupu;Morris Sloman

  • Affiliations:
  • -;-;-;-

  • Venue:
  • ICAS '09 Proceedings of the 2009 Fifth International Conference on Autonomic and Autonomous Systems
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Policies form an important part of management and can be an effective means of implementing self-adaptation in pervasive systems. Most policy-based systems focus on large-scale networks and distributed systems. Consequently, they are often fragmented, dependent on infrastructure and lacking flexibility and extensibility. This paper presents Pon- der2, a novel policy system that is suitable for a wide range of environments and applications. The design and implementation of Ponder2 emphasises simplicity, flexibil- ity and extensibility and provides users with the ability to interact easily with the managed system. Ponder2 can interact with other software and hardware components and is being used in environments ranging from single devices, to personal area networks, ad-hoc networks and distributed systems. We also describe PonderTalk, a high-level object orientated language inspired by Smalltalk for configuring and controlling Ponder2 systems.large scale systems. It advocates a similar de-centralised model of autonomous agents co-operating with each other and composing into more complex configurations. However, many existing policy-based frameworks have not been conceived for such environments. Their design is dependent on centralised infrastructure support such as LDAP directories and CIM repositories. Their deploy- ment model is often based on centralised provisioning and decision-making that does not offer the means for policy execution components to interact with each other, collabo- rate or federate into larger structure. Policy specification is seen as an off-line activity, and policy frameworks do not easily interact with the managed system. Consequently such frameworks are difficult to install, run and experiment with. Additionally, they usually do not scale to smaller devices omnipresent in pervasive systems.