Infrastructure automation based on prediction

OData support
Supervisor:
Dr. Szatmári Zoltán
Department of Measurement and Information Systems

Nowadays, IT services are being used more and more widely. The load generated by them can differ in time due to the number of the requests received and the nature of the services. In order to ensure the quality of services, this fluctuation needs to be handled for the optimal, seamless user experience.

It is advisable to react to the fluctuation of the load by varying the amount of resources. For the optimal solution, the services need to be decomposed into multiple smaller sub-services by implementing a Microservice Architecture. Using this infrastructure only the appropriate components are necessary to be duplicated. With the spread of the cloud-based technologies, dynamic scaling of resources can be ensured, while only the actually used resources are shown as costs.

Sometimes it is not enough to respond to the changing loads by monitoring actual re-source requirements. In order to provide more predictable and stable service, the auto-mation of the infrastructure should be based on prediction bases, in which case we pre-pare for the actual load taking into account past measurements.

In my thesis I implement and present my experimental dynamic infrastructure and the required background knowledge. The infrastructure is supported with automatic service discovery and load balancing, I demonstrate the automation process the simulated load through an example. I also present, how future loads can be estimated with the predic-tion method developed based on past observations.

Downloads

Please sign in to download the files of this thesis.