Chiron is a serverless deployment system that implements wrap, a novel abstraction for the "m-to-n" serverless deployment model aiming at optimizing performance and resource efficiency by function partition and function execution mode selection, atop OpenFaaS with combined processes and threads.
We implement the Chiron system in OpenFaas v0.21.1, and run it on a Kubernetes v1.23.6 cluster. The Installation shows the instruction to install the environment.
We provide some benchmark for evaluating our system. The Benchmark shows the instruction to build and deploy several serverless workflows.
This module shows how Chiron profiles a function for further performance prediction. The details can be found in Profiler.
This module is core of Chiron, which determines the wraps of workflow based on the given SLO. The details can be found in Scheduler.
This module produces the orchestrator code of wraps based on the result of Scheduler. The details can be found in Generator.
This sections contains a simple instruction for replaying our evaluation. The details can be found in Envaluation.
Welcome to cite Chiron in SC'23 by:
@inproceedings{10.1145/3581784.3613211,
author = {Li, Yiming and Zhao, Laiping and Yang, Yanan and Qu, Wenyu},
title = {Rethinking Deployment for Serverless Functions: A Performance-First Perspective},
year = {2023},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3581784.3613211},
doi = {10.1145/3581784.3613211},
booktitle = {Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis},
numpages = {14},
location = {Denver, CO, USA},
series = {SC '23}
}