Leveraging Machine Learning to Reduce Cold Start Latency of Containers in Serverless Computing

Bannon, Ryan (2022) Leveraging Machine Learning to Reduce Cold Start Latency of Containers in Serverless Computing. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration manual) Download (1MB) | Preview

Abstract

Leading cloud service providers offer Function-as-a-Service (FaaS) which allows users to outsource the provisioning and management of servers and focus solely on the business logic that drives their operations. This means users can run code in stateless compute containers living in an event driven ecosystem which is fully managed by the cloud provider. Therefore, this model masks the underlying infrastructure and implementation stack that these services are built upon. The core principles of the microservices approach include decentralisation, fault tolerance, continuous delivery, and deployment. However, decentralisation within an eventdriven domain often results in chain reaction. Functions invoke services which call other function and so on. As such, one major drawback in serverless computing is a problem called cold start. The nature of these services forces the scaling components to warm, freeze, thaw and terminate container state, often times at the expense of latency upon subsequent invocations. This paper address this problem with a recent iteration of the LSTM neural network called Gated Recurrent Unit (GRU). Results from experiments that simulated real Azure Functions traces, recorded improvements of up to 28% with its function execution predictions and cold start solution.

Item Type:	Thesis (Masters)
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Cloud computing Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Cloud Computing
Depositing User:	Tamara Malone
Date Deposited:	06 Dec 2022 15:20
Last Modified:	08 Mar 2023 15:29
URI:	https://norma.ncirl.ie/id/eprint/5971

Actions (login required)

View Item