FaaS Scheduling Benchmark

This benchmark suite is designed for evaluating the scheduler in a Function-as-a-Service platform. It consists of functions from various application domains so as to simulate the mixed workload in production.

Prerequisite

If you want to deploy and run the functions, you need to ...

1. ... have an Alibaba Cloud account because the functions are implemented for Function Compute.

2. ... know how to use fun CLI tool. If not, you could check out this Get Started guide.

Repo Structure

The benchmark suite has five applications, each with two functions. You could follow the README in each application directory to deploy and invoke the function.

Application	Function	Programming Language	Dependencies
Smart Parking	Query Vacancy	JavaScript	Redis
	Reserve Spot	JavaScript	Redis, Kafka
Log Processing	Anonymize Log	Rust	Kafka
	Filter Log	Rust	Kafka
Computer Vision	Detect Object	Python	TensorFlow
	Classify Image	Python	TensorFlow
Media Processing	Get Media Meta	Python	OSS
	Convert Audio	Python	OSS
Smart Manufacturing	Ingest Data	C++	MySQL
	Detect Anomaly	C++	MySQL

Portability

The benchmark suite only supports Function Compute. To port it to other FaaS platforms, you have to change the way how arguments are passed into the function and how the functions are deployed.

Citation

If you have used this benchmark in your research project, please cite the following paper.

@inproceedings{tian2022owl,
  title={Owl: Performance-Aware Scheduling for Resource-Efficient Function-as-a-Service Cloud},
  author={Tian, Huangshi and Li, Suyi and Wang, Ao and Wang, Wei and Wu, Tianlong and Yang, Haoran},
  booktitle={Proceedings of the ACM Symposium on Cloud Computing 2022},
  year={2022}
}

Acknowledgement

@SimonZYC has contributed to this benchmark.