A Cost-effective auto-scaling strategy between server and serverless architectures for cloud deployments

Abstract

Nowadays, cloud computing is marking as an emerging technology in the universe. Most of people use cloud environments to deploy their systems. Many enterprise systems give their benchmarks according to the specific instances in the cloud. Cloud providers provide primary services as SaaS, PaaS, IaaS and FaaS, then Cloud consumers use those services as per their demands. The systems which are developed on several years ago are still running upon on-premise environments. They are tightly coupled legacy systems. Hence people cannot adopt new technologies to those monolithic systems. Nevertheless, people who are in the software industry need to integrate new technologies into their systems to stay competitive in the IT industry. To get into the new technology that should be cost-effective, more reliable, and should need to cater the new and existing functionalities of the system. One of the problems is that migrating the whole system or partial system to the new technology is hazardous. Sometimes the system needs to handle substantial traffic loads, according to the business needs. In order to satisfy sudden traffic spikes, systems need to scale horizontally or vertically. If their application server is deployed upon the on-premise server, scaling is complicated, and if the application is deployed on the cloud, VMs scaling is possible, but it can be very costly. In this research, the goal is to overcome these scaling problems. The research component presents the strategy to load balancing the traffic between the server and serverless function. The cost of the serverless function calculated only for the executes requests, so it is cheaper than running a new server. Therefore this hybrid strategy is a highly cost-effective way to scaling the servers and also proven that the proposed system is capable of load balancing the traffic according to the server loads with the analytical results. Other researches are proven that this research component is the right solution for ongoing industry problems.