Design and Validation of an Analytical Model to Evaluate Monitoring Frameworks Limits

Abstract

It is  essential that  a  monitoring  system is  being designed   with performance and  scalability  in mind.   But due  to the diversity and complexity  of both the  monitoring and  the  monitored systems, it is currently difficult to    reason on both   performance and scalability using  ad hoc   techniques.  Thus, both   simulation  is required  and analytical models based   on   well established  techniques  such    as queueing theory have to be developed. In  this paper we provide an analytical    modeling of  the     behavior  of  the    commonly used manager-agent   monitoring frameworks   within  two scenarios:  single manager-single agent  and   single manager-multiple agents.   The two designed   models  enable the  automation  of   the estimation of  the scalability limit  of the two types  of management  monitoring schemes regarding a performance metric like the monitoring delay.  We validate our developed  models  through simulation  based on  parameters values obtained from the performance measurement  of a life running JMX-based monitoring applications with the two scenarios.