Cost optimization for I/O intensive workload in FaaS

Singh, Utkarsh Kumar (2024) Cost optimization for I/O intensive workload in FaaS. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (2MB) | Preview

Abstract

FaaS is the most popular software paradigm in cloud computing world, and it’s adoption is increasing day by day. FaaS provides a lucrative option for companies, government and individual developers as it is based on pay as you go model instead of traditional model where customers has to reserve the resource for specific period irrespective of whether the resource is idle or in use. FaaS billing model is based on function execution time, memory and CPU usage. This model works well for CPU computation workload whereas it fails when workload is I/O intensive where CPU and memory are not utilized fully thus leading to unfair charges to customer for these underutilized resources. In this project an framework is proposed where workload is divided into two parts, where first one is CPU intensive workload and other is I/O intensive workload which are executed separately. Amazon Kinesis is leveraged for its scalability, deduplication and high throughput for dynamically adjusting resources of lambda function with the help of machine learning to predict memory and timeout of lambda function. AWS Cloudwatch is used to monitor timeout of lambda function if lambda function is timeout, alarm is triggered and model is trained again. The proposed framework has saved cost by 16%-32% for specific type of I/O intensive workload.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Jaswal, Shivani UNSPECIFIED
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
Divisions:	School of Computing > Master of Science in Cloud Computing
Depositing User:	Ciara O'Brien
Date Deposited:	17 Jul 2025 12:27
Last Modified:	17 Jul 2025 12:27
URI:	https://norma.ncirl.ie/id/eprint/8157

Actions (login required)

View Item