Serverless in the Wild: Characterizing and Optimizing the Serverless Workload at a Large Cloud Provider

Function as a Service (FaaS) has gained tremendous popularity as a way to deploy computations to serverless back ends in the cloud. We performed the first characterization of an entire production FaaS environment (Azure Functions). Our characterization revealed many unique aspects of serverless workloads compared to traditional cloud applications. Using this deep understanding, we designed a new dynamic resource management policy to improve the performance and reduce the memory footprint of serverless workloads. This new policy is now deployed in production, and our characterization data traces are publicly released for researchers. Our paper won the Community Award at USENIX ATC 2020.

Related talks:

Title: Serverless in the Wild: Characterizing and Optimizing the Serverless Workload at a Large Cloud Provider [Slides] [Video]
Speakers: Rodrigo Fonseca and Mohammad Shahrad
Venue: USENIX ATC '20
Date: July 15, 2020
Title: Serverless in Seattle: Toward Making Serverless the Future of the Cloud [Video]
Speakers: Ricardo Bianchini
Venue: SYSTOR '20 Keynote
Date: October 14, 2020

Our paper has received some press coverage, including:

Rodrigo Fonseca Wins The USENIX Community Award For Characterizing And Optimizing A Large Serverless Workload. Brown Computer Science. July 21, 2020.
Read a paper: Serverless in the wild. Vivek Haldar. July 6, 2020.
Eliminate Cold Starts by Predicting Invocations of Serverless Functions. Mikhail Shilkov. June 18, 2020.
Serverless in the Wild: Azure Functions Production Usage Statistics. Mikhail Shilkov. May 5, 2020.
Serverless in the wild: characterizing and optimising the serverless workload at a large cloud provider. The Morning Paper. June 18, 2020.

We are also honored to see our paper being features in many advanced computing courses.