http://lattes.cnpq.br/9548167960831148; GOES, Armstrong Mardilson da Silva.
Résumé:
In the last decade, a large growth on the usage of virtualized infrastructures for application
deployment has been observed. Seeking flexibility on resources allocation and lower management cost, many companies adopted computing clouds as environments to perform some of their crucial activities. A problem observed in such environments is the assurance of quality of service in the execution of applications. The amount of resources required by applications varies with the time. Peaks on the amount of requests in executions of interactive applications, such as e-commerce websites, are common and is necessary to add resources in order to keep a reasonable response time. In executions of batch applications, normally we want to ensure a deadline, related to the application’s priority. Estimating the required amount of resources to assure execution deadline is difficult and prone to underallocations or superallocations. A different solution is to modify the amount of allocated resources in case the progress is not satisfactory. Following this strategy, this work proposes a provisioning method based on PID controllers and vertical scaling. In order to evaluate the proposed provisioning method on assuring quality of service of batch applications, we executed two microbenchmarks and one application used in Big Data, on a real cloud infrastructure where the provisioning was controlled by our method. We collected the execution times and resources usage data from the microbenchmarks and the test application. Results show that the provisioning method is feasible and effective on ensuring quality of service, although it needs application-dependent configuration.