Performance modeling of cloud apps using message queueing as a service (MaaS)

This paper presents an analytical model to study the performance of cloud applications using message queueing as a service (MaaS). MaaS is a cloud service which allows the development departments to focus on delivering business and computing applications without being concerned with the underlying message queueing infrastructure to be scalable, secure, and reliable. Estimating the service delay (prior to provisioning cloud resources) of this type of cloud apps is an important engineering and resource management problem. Such estimation would help in computing the overall network and service delay that users experience. In a way, cloud providers would allocate the appropriate capacity for the needed cloud resources to meet the Service Level Agreement (SLA) parameters. In this paper, we present an analytical model by using Markov chain to study the performance of cloud apps which use MaaS. Given the expected request arrival rate, the queue size, and the expected service rate of each processing stage of the cloud app, our analytical model can estimate the app performance in terms of key SLA parameters which include response time, throughput, and request loss. In addition, our model yields equations for other key performance measures which include system idleness and utilization, queuing delay, and system and queue occupancies. Our analytical model is verified and validated by using discrete-event simulation and experimental measurements taken from an experiment conducted on AWS (Amazon Web Services) cloud.