Bridging the Sustainability Gap in Serverless through Observability and Carbon-Aware Pricing

Changyuan Lin (University of British Columbia); Mohammad Shahrad (University of British Columbia)

Abstract

Serverless computing has become a mainstream cloud computing paradigm due to its high scalability, ease of server management, and cost-effectiveness. With cloud data centers’ carbon footprint rising sharply, understanding and minimizing the carbon impact of serverless functions becomes crucial. The unique characteristics of serverless functions, such as event-driven invocation, pay-as-you-go billing model, short execution duration, ephemeral runtime, and opaque underlying infrastructure, pose challenges in effective carbon metering. In this paper, we argue that the current carbon estimation methodologies should be expanded for more accurate carbon accounting in serverless settings, and propose a usage and allocation-based carbon model that aligns with the context of serverless computing. We also articulate how current serverless systems and billing models do not make it financially attractive to prioritize sustainability for a broad class of users and developers. To solve this, we propose a new carbon-aware pricing model and evaluate its ability to incentivize sustainable practices for developers through better alignment of billing and carbon efficiency.