The Impact of Auto-Refactoring Code Smells on the Resource Utilization of Cloud Software
Cloud-based software-as-a-service (SaaS) have gained popularity due to their low cost and elasticity. However, like other software, SaaS applications suffer from code smells, which can drastically affect functionality and resource usage. Code smell is any design in the source code that indicates a deeper problem. The software community deploys automated refactoring to eliminate smells which can improve performance and also decrease the usage of critical resources. However, studies that analyze the impact of automatic refactoring smells in SaaS on resources such as CPU and memory have been conducted to a limited extent. Here, we aim to fill that gap and study the impact on resource usage of SaaS applications due to automatic refactoring of seven classic code smells: god class, feature envy, type checking, cyclic dependency, shotgun surgery, god method, and spaghetti code. We specified six real-life SaaS applications from Github called Zimbra, OneDataShare, GraphHopper, Hadoop, JENA, and JAMES which ran on Openstack cloud. Results show that refactoring smells by tools like JDeodrant and JSparrow have widely varying impacts on the CPU and memory consumption of the tested applications based on the type of smell refactored. We present the resource utilization impact of each smell and also discuss the potential reasons leading to that effect.