PSElastic : um modelo de elasticidade multinível para Brokers Publish/Subscribe

Abstract

The Internet of Things (IoT) is considered the next big evolution in the computing era after the Internet. In Internet of Things many objects around us will be connected in networks and will communicate among themselves without assistance or human intervention. Large scale implementations has potential to create a huge aggregate traffic. Often brokers or intermediaries, are used to enable this machine-to-machine (M2M) communication using the Publish/Subscribe communication model that has several private libraries for use in Internet of Things. Some works developed by the community aim to treat performance in Publish/Subscribe brokers environment, however impose modifications and restrictions regarding brokers and architecture used. In this scenario, aiming to fulfill the need for cited scalability, we propose a elasticity model called PSElastic. The model provides an model with multilevel elasticity for Publish/Subscribe Brokers, making use of the elasticity promoted by cloud computing, not affecting system performance and without the need for investments or adjustments by customers and applications that make use of this model. PSElastic offers elasticity in the following manners: automatic, reactive and uses the multilevel elasticity, where multilevel refers to Broker level and Orchestrator model. In order to evaluate the model PSElastic, three different scenarios for execution of tests were defined: Scenario 1, without elasticity; Scenario 2, elasticity in the layer of Broker and Scenario 3, multilevel elasticity. Different limits of replica of virtual machines were used, as well as thresholds to evaluate Execution Time, number of Messages per Second handled and Efficiency in the use of computational resources. PSElastic showed excellent results by reducing the time needed to handle a data load by up to 81.2% and increase the throughput of the environment with Mosquitto Broker from 54.95 messages per second (Mps) to 235.29 Mps and from 38.02 Mps to 202.02 Mps in the environment with RabbitMQ Broker.