The Case for Time-Shared Computing Resources

TL;DR

This paper argues for a paradigm shift towards higher-level resource sharing in computing infrastructure, advocating fewer physical resources with acceptable performance trade-offs to improve energy efficiency and sustainability.

Contribution

It introduces a novel perspective on resource management, emphasizing higher abstraction sharing to reduce physical resource needs and enhance energy efficiency.

Findings

Time-sharing is declining due to performance and cost reasons.

Higher-level resource sharing can lead to energy savings.

Fewer physical resources with managed performance trade-offs are feasible.

Abstract

The environmental impact of Information and Communication Technologies (ICT) continues to grow, driven notably by increasing usage, rebound effects, and emerging demands. However, despite the virtual nature of its services, the sector remains inherently constrained by its materiality and cannot rely on an infinite pool of resources. As a result, the wide variety of supported services may need to be managed under stricter limits within hosting facilities in the future. Contrary to common assumptions, we show that tenants typically do not share computing resources, even in environments commonly perceived as mutualized, such as cloud platforms. Time-sharing has been progressively phased out for reasons of performance, security, predictability, and, perhaps more importantly, due to the decreasing cost of computing resources. This paper advocates for managing fewer physical resources by improving resource sharing between tenants. It represents a paradigm shift, moving beyond traditional time-sharing at the hardware level to a higher abstraction. This approach entails "doing with fewer resources" under conditions of "reduced performance". Nonetheless, enhancing the mutualization of infrastructure can reduce cluster sizes (through consolidation) and improve energy efficiency, with gains related to the accepted performance trade-off, a situation potentially more socially acceptable than eliminating services. We review the current state of the art, identify challenges and opportunities, propose interpretations of Time-Shared Computing, and outline key research directions.