Towards QoS-Aware Function Composition Scheduling in Apache OpenWhisk

G. Russo Russo, A. Milani, V. Cardellini, S. Iannucci

Proc. of *LESS '22 (co-located with PerCom '22)

[doi]

Function-as-a-Service (FaaS) is increasingly popular thanks to the benefits provided to application developers and operators. Besides commercial Cloud-based offerings, open-source solutions have emerged enabling FaaS deployment on private infrastructures and possibly at the edge of the network. When moving from the Cloud to Fog/Edge environments, optimizing resource allocation for function execution becomes a critical challenge. Unfortunately, existing FaaS platforms have little or no support for fine-grained scheduling and resource allocation, nor allow users to enforce Quality-of-Service (QoS) requirements. We take a first step towards the development of a QoS-aware FaaS platform. We design and implement new mechanisms to support differentiated classes of services within Apache OpenWhisk, a popular open-source FaaS framework. Our experiments show that our protopublication_type efficiently supports state-of-the-art scheduling policies and provides throughput improvements when dealing with function compositions under high load scenarios.