🤖 AI Summary
To address network performance bottlenecks arising from surging datacenter traffic and the slowing of Moore’s Law, this paper proposes the Reconfigurable Datacenter Network (RDCN) architecture—a paradigm shift from static topologies. We introduce the first systematic taxonomy of RDCNs (categorized along static/dynamic and oblivious/aware dimensions) and establish a formal model that uncovers causal relationships between spatiotemporal traffic patterns and topology evolution. Integrating optical circuit switching, dynamic graph modeling, traffic demand forecasting, and topology optimization algorithms, RDCN enables on-demand, adaptive, real-time reconfiguration. Experimental evaluation demonstrates that demand-aware reconfiguration reduces tail latency by over 40% and improves throughput utilization by 2–3×. This work establishes a new, highly elastic, energy-efficient, and scalable network paradigm for next-generation datacenters—one that actively adapts to dynamic workloads rather than passively accommodating them.
📝 Abstract
With the popularity of cloud computing and data-intensive applications such as machine learning, datacenter networks have become a critical infrastructure for our digital society. Given the explosive growth of datacenter traffic and the slowdown of Moore's law, significant efforts have been made to improve datacenter network performance over the last decade. A particularly innovative solution is reconfigurable datacenter networks (RDCNs): datacenter networks whose topologies dynamically change over time, in either a demand-oblivious or a demand-aware manner. Such dynamic topologies are enabled by recent optical switching technologies and stand in stark contrast to state-of-the-art datacenter network topologies, which are fixed and oblivious to the actual traffic demand. In particular, reconfigurable demand-aware and 'self-adjusting' datacenter networks are motivated empirically by the significant spatial and temporal structures observed in datacenter communication traffic. This paper presents an overview of reconfigurable datacenter networks. In particular, we discuss the motivation for such reconfigurable architectures, review the technological enablers, and present a taxonomy that classifies the design space into two dimensions: static vs. dynamic and demand-oblivious vs. demand-aware. We further present a formal model and discuss related research challenges. Our article comes with complementary video interviews in which three leading experts, Manya Ghobadi, Amin Vahdat, and George Papen, share with us their perspectives on reconfigurable datacenter networks.