The Impact of Database Layer on Auto-Scaling Decisions in a 3-Tier Web Services Cloud Resource Provisioning

Abstract

This paper investigates the impact of the database layer on the scaling actions of the business layer of a 3-tier web service system in cloud resource provisioning. The research question is "What is the impact of the database layer on the business layer auto-scaling decisions?" In this work two hypotheses are tested: 1) "Database tier capacity has no effect on the business tier scaling decisions" and 2) "Scaling up of a database tier increases Service Level Agreement (SLA) violations." To test the hypotheses, an auto-scaling simulation package based on Queuing Network Models (QNM) and Layered Queuing Network Models (LQNM) is developed. The auto-scaling simulation package is used to investigate the database impact on the business tier scaling decisions in the cloud environments with three different workload patterns (growing, periodic, and unpredictable patterns). This paper also provides an analytical investigation that empirically validate the hypotheses. The results suggest that the database tier has no effect on the business tier scaling decisions. However, decreasing the capacity of the database layer increases the rate of the SLA violations.