Monitoring multiple concurrent service level parameters with multidimensional trees

Abstract

The introduction of new computing paradigms in the Internet as well as the increasing size and complexity of services and resources demand the development of new approaches for defining and monitoring service levels. It is often necessary to keep track of multiple concurrent service level requirements. In this paper we present a service level monitoring strategy that allows both online and offline tracking the performance of multiple concurrent resources. Data is collected with SNMP (Simple Network Management Protocol). The strategy is based on building multidimensional search trees. k-d (k-dimensional) trees are employed for online continuous monitoring, and k-d-B trees are employed for offline monitoring, based on logs of monitored data. Searching with the proposed strategy has cost O(logN) where N is the number of samplings or log size. The strategy allows clients and providers to confirm whether contract specifications were hold or not, and for how long. Experimental results are presented, including a comparison of the proposed approach with a traditional database. A practical tool was implemented and results are shown for a set of monitored Web and Video servers, as well as for monitoring data obtained from a real Telecom billing system. © 2009 Springer-Verlag.