Cloud Functions for Fast and Robust Resource Auto-Scaling

Download PDF
Open Webpage

Joe H. Novak, Sneha Kumar Kasera, Ryan Stutsman

Published: 2019, Last Modified: 28 Jan 2025COMSNETS 2019EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: We design and build FEAT, a new scaling approach that uses (1) cloud functions as interim processing resources to compensate for VM launch delays and (2) a reactive, knobless, auto-scaling algorithm that requires no pre-specified thresholds or parameters, making it robust against changing load. We implement FEAT on Amazon Web Services (AWS) and Microsoft Azure. Our evaluations clearly demonstrate the higher performance and robustness of FEAT in comparison to existing approaches.