pacslab/serverless-temporal-perf-modeling

Temporal Performance Modelling for Serverless Computing Platforms.

Temporal Performance Modelling of Serverless Computing Platforms

This repository includes all artifacts for our paper "Temporal Performance Modelling of Serverless Computing Platforms" presented in the Sixth International Workshop on Serverless Computing (WoSC6) 2020 . The performance model presented in our work is capable of predicting several key performance indicators of serverless computing platforms, while maintaining fidelity and tractabality thoughout the parameter space.

Benefits

• Works with any service time distribution (general distribution).
• Predicts transient characteristics, making it a proper candidate for use in serverless computing management systems.
• Is tractable while having a high fidelity.

Artifacts

• Deployment code for collecting data using experimentations
• Code for parsing the datasets and generating plots

Requirements

• Python 3.6+
• PIP

Installation

pip install -r requirements.txt

License

Unless otherwise specified:

MIT (c) 2020 Nima Mahmoudi & Hamzeh Khazaei

Citation

You can find the paper with details of the proposed model in PACS lab website. You can use the following bibtex entry:

@inproceedings{mahmoudi2020tempperf,
  author={Mahmoudi, Nima and Khazaei, Hamzeh},
  title={{Temporal Performance Modelling of Serverless Computing Platforms}},
  year={2020},
  publisher = {Association for Computing Machinery},
  booktitle={{Proceedings of the 6th International Workshop on Serverless Computing}},
  pages={1--6},
  numpages = {6},
  keywords = {performance modelling, serverless computing, serverless, temporal, transient, performance},
  location = {TU Delft, Netherlands},
  series = {WOSC '20}
}