The Case for Automated Planning in Autonomic Computing

Srivastava, Biplav; Kambhampati, Subbarao

doi:10.1109/icac.2005.58

Cited by 12 publications

(9 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Policy-based management has been an active area of research for the past few years and many projects have focused on designing policy languages [6,13,15] detecting and resolving policy conflicts [2,8,11] and various other analyses [7,12,20]. To the best of our knowledge, no research work on policy-based management has addressed the problem of ordering management rules and providing enforcement guarantees as we have addressed in this paper.…”

Section: Related Workmentioning

confidence: 93%

Ordering management actions in pervasive systems using specification-enhanced policies

Chankar

Campbell

2006

Fourth Annual IEEE International Conference on Pervasive Computing and Communications (PERCOM'06)

View full text Add to dashboard Cite

A pervasive system features a plethora of devices, services and applications organized as a large distributed system. One approach to managing such systems is by policies where administrators specify the management action to be taken in different situations using Event-Condition-Action (ECA) rules. An important problem with policy-based management of a pervasive system is that multiple rules can get triggered on a single event and the behavior of the system depends on the order of rule enforcement. Systems managed using ECA policies do not provide guarantees about system behavior when multiple rules are concurrently triggered.In this paper, we present a novel rule framework called Event-Condition-Precondition-ActionPostcondition (ECPAP) that combines axiomatic specifications with ECA rules for specifying management rules. ECPAP rules contain action specifications in first-order predicate logic that enables us to reason about the enforcement order. We define a notion called enforcement semantics for policy-based management and show how this can be used to provide guarantees about system behavior. We present the details of the framework.

show abstract

Section: Related Workmentioning

confidence: 93%

Ordering management actions in pervasive systems using specification-enhanced policies

Chankar

Campbell

2006

Fourth Annual IEEE International Conference on Pervasive Computing and Communications (PERCOM'06)

View full text Add to dashboard Cite

show abstract

“…fault) occurs, the self-healing is the problem of finding an action sequence to recover the element to the previous or a new safe state by minimum cost, providing the above six entities and H as history of actions. According to [5] planning has the best potential of use in self-healing to find the minimum expected cost for actions through continuous planning. Similarly, self-optimizing is the problem of finding an action sequence to optimize U (.)…”

Section: Coordination Mechanismmentioning

confidence: 99%

A coordination mechanism for self-healing and self-optimizing disciplines

Salehie

Tahvildari

2006

Proceedings of the 2006 International Workshop on Self-Adaptation and Self-Managing Systems

View full text Add to dashboard Cite

There is an increasing demand for autonomic systems, which offer controlling complexity through a decentralized, multidiscipline and policy-based paradigm. In practice, only one discipline is often taken into account, while having shared resources and policies, it is required to coordinate different disciplines. This research addresses the problem of coordinating self-healing and self-optimizing in autonomic elements by generic modelling of disciplines, and proposing a coordination mechanism.

show abstract

“…A notable point is that planning, at least in its basic form, cannot be used for all of the self-* properties. According to [197], among all of the self-* properties, planning has the highest potential for being used in conjunction with self-healing. One example of using AI-planning for self-healing is the work by Arshad et al [11].…”

Section: Supporting Artificial Intelligence Conceptsmentioning

confidence: 99%

“…Technically, it is possible to design the mechanism with the aid of AI planning algorithms like STRIPS [133]. However, as pointed out by other researchers, planning is not an suitable option for all self-* properties-such as self-optimizing-and their associated quality goals [197]. Moreover, in run-time adaptation, because the adaptation manager is receiving feedback continuously from the external world, continuous planning is required.…”

Section: Goal-centric Mechanismsmentioning

confidence: 99%