BEHAVE – Behavioral Modeling of Application Functions in Serverless Computing
Name of the participant: Anshul Jindal
Description of the IT-research project: The automatic management of resources in serverless computing facilitates application development by shifting the burden to the cloud platform. However, already existing challenges like the fast startup of containers, communication, and latency of data accesses are further increased due to the heterogeneity in the computing continuum (edge-cloud continuum). On the other side, heterogeneity offers a wide range of opportunities for meeting different objectives like SLO requirements and energy efficiency in unconventional ways. Therefore, resource management decisions like function scheduling and data placement have to take into account all those challenges.
At the Chair for Computer Architecture and Parallel Systems, TU München, we are working on the extension of FaaS for heterogeneous platforms (the computing continuum) using a network of distributed heterogeneous target platforms called Function Delivery Network (FDN) analogous to Content Delivery Network. FDN provides Function Delivery as a Service (FDaaS), delivering the function to the right target platform based on the required computational and data demand.
The goal of this project is to enable autonomous decision making for FDN by characterizing functions in FaaS applications via various behavioral models. The project will focus on the following management use cases of the FDN:
- Application function scheduling: Scheduling functions to the heterogeneous resources based on the function’s cost, response time, and computation and data requirements.
- Data storage and management: Placement of data and data migration required by functions.
- Operation and Maintenance: Distributed anomaly detection algorithms for supporting smart O & M in computing continuum.
Implementation period: 01.02.2021 – 31.10.2022