Benchmark and Simulation of Cloud Functions (FaaS)
Function as a Service (FaaS) is the predominant reason why a lot of researchers and practioneers talk about Serverless Computing. Since Serverless is misleading per se, we use the terms cloud function when talking about concrete instances of a deployed function and FaaS when talking about the concept to deal with the concept of short lived function running ephemerally in the cloud.
The goal of this PhD project is to understand runtime characteristics of the platform, function characteristics of the deployed cloud function and take dependent services like database into consideration when building a local clone of the platform at a developer's machine. These aspects enable us to configure the cloud function appropriately to the specified requirements upfront. Furthermore, we are able to build a simulation and benchmarking tool to conduct repeatable and fair experiments which is the second part of the mentioned project.
Publications:
- Overall Idea: Towards Performance and Cost Simulation in Function as a Service
(ZEUS 2019) - Cold Start Influencing Factors in Function as a Service
(WoSC 2018, UCC Companion) (v0.1 of SeMoDe) - Troubeshooting Serverless Functions: A Combined Monitoring and Debugging Approach
(SummerSoC 2018) - Impact of Application Load in Function as a Service
(SummerSoC 2019) - Optimizing Cloud Function Configuration via Local Simulations
(CLOUD 2021) (v0.3 of SeMoDe) - Why Many Benchmarks Might Be Compromised
(SOSE 2021) (v0.4 of SeMoDe) - SeMoDe – Simulation and Benchmarking Pipeline for Function as a Service
(Report, November 2021) (v1.1 of SeMoDe) - Resource Scaling Strategies for Open-Source FaaS Platforms compared to Commercial Cloud Offerings
(CLOUD 2022) (v1.2 of SeMoDe) - A Structured Literature Review Approach to Define Serverless Computing and Function as a Service
(CLOUD 2023)
Prototype:
SeMoDe - Serverless Monitoring and Debugging
This prototype implements the concepts we specified in the paper and tries to form the mentioned benchmarking and simulation tool at the end of the PhD project.
Participants: