Scalable Markov Chain Based Algorithm for Fault-Isolation in Autonomic Networks

Abstract

Research in the area of Autonomic Networks is on the rise. Autonomicity-realized through control loops, is an enabler for advanced self-manageability of network nodes and devices. Therefore, the specification and design of autonomic behaviors is required for each of the management functions defined by the well established FCAPS network management framework (Fault-, Configuration-, Accounting-, Performanceand Security-Management). In the context of Autonomic FaultManagement, mechanisms and algorithms are required that enable efficient and scalable interactions among the FaultManagement processes defined by the TMN (Telecommunications Management Network) standard. TMN defines Fault-Detection, Fault-Isolation, and Fault-Removal as the processes involved in Fault-Management. Therefore, in Autonomic Networks, some capabilities of Fault-Isolation must be in-built into a node and into the whole fundamental network architecture (apart from those aspects handled in the management plane), and its results must be fed into the embedded automatic Fault-Removal mechanisms. This imposes some scalability requirements on the employed algorithms. In this paper, we propose a novel scalable Markov Chain based algorithm for on-line Fault-Isolation. Furthermore, we analyze its computational and space complexity and evaluate its fault identification capabilities, as well as scaling properties on potential issues within an IPv6 network.

Scalable and fast root cause analysis using inter cluster inference

“…Then, several methods have been proposed in the literature in order to design more powerful fault diagnosis solutions. These proposals are based on graph theory and artificial intelligence tools, such as: Marckov Chains [11], Causal Graphs [7], Bayesian Networks [13], [8], Codebook [14] and Rule-Based Reasoning [10]. They are based on the construction, a priori, of a dependency model between all the metrics characterizing the equipments of an operator network [7], [12], [13].…”

Framework for distributed autonomic self‐healing in fixed IPv6 networks

“…These experiences have led to the identification of new requirements and a redesign of the framework as presented in this article. Moreover, a Fault‐Isolation algorithm was specially developed for the goals of our framework. This algorithm is used as a library within the implementation of the prototype presented here and correspondingly plays a role for the reported numerical results.…”

Robust architecture for distributed intelligence in an IP-based mobile wide-area surveillance system

“…Further information on the algorithms implemented in the above described components, as well as on the case studies, can be found in [35][36][37][38][39].…”

“…Thereby, the FDLI functions use a causality/fault-propagation model that is stored in the belonging repository. For traversing such a fault-propagation, we use a Markov Chain based algorithm that we specified and analyzed in our previous work [35]. Thereby, we systematically infer the probability for particular root causes to be the explanation of the experienced symptoms by combining insights from the areas of probability theory, linear algebra and data structures' design.…”

“…This component should be referred by both DEs to synchronize and eventually allow or disallow their tentative actions. In the course of that, the ASF use a utility function-based approach [35][36][37] to synchronize the tentative actions and their potential impacts on the system. Within this approach, an optimization problem is solved for the set of tentative actions to select the best actions that do not contradict and improve the overall health of the system.…”