High Level Description

A common problem in high traffic systems is scaling. When applications take a long time to initialize, scaling in the middle of a traffic spike can cause a noticeable latency impact and even causing downtime. Predictive scaling together with machine learning involves gathering and utilizing data to accurately forecast incoming spikes and efficiently allocate resources within a microservices-based infrastructure. By using machine learning, systems can predict the upcoming traffic and scale the resources in real-time, optimizing performance and minimizing downtime. This approach enables businesses to stay ahead of their resource needs, ensuring seamless and cost-effective operation of their cloud-based services.

Project Description

In this thesis, design and implement predictive scaling in a cloud microservice architecture using machine learning. The goal is to build a model that can accurately forecast spikes and scale the system accordingly, based on gathered data of traffic patterns such as CPU utilization, memory usage and other metrics. Compare your solution by implementing and/or analyzing other scaling methods, such as reactive and proactive scaling.

Who are we looking for?

Bachelor/Master of Science in Computer Science/Engineering

Purpose

In this thesis, investigate these questions:

• How well does the solution perform in comparison to other scaling methods
• What are the benefits when it comes to resource optimization and cost efficiancy
• How much data is required to train the machine learning model
• What other metrics can be used to improve the machine learning model