Reconfigurable datacenter networks use fast optical circuit switches to provide high bandwidths at low cost, therefore emerging as a compelling alternative to packet switching. These switches offer micro- and nano-second reconfiguration, and reacting to demand at this time scale is infeasible. Proposed designs have therefore largely been oblivious, supporting arbitrary traffic patterns. However, this imposes a fundamental latency-throughput tradeoff that significantly limits the benefits of these switches.

Circuit-switched technologies have long been proposed for handling high-throughput traffic in datacenter networks, but recent developments in nanosecond-scale reconfiguration have created the enticing possibility of handling low-latency traffic as well. The novel Oblivious Reconfigurable Network (ORN) design paradigm promises to deliver on this possibility. Prior work in ORN designs achieved latencies that scale linearly with system size, making them unsuitable for large-scale deployments. Recent theoretical work showed that ORNs can achieve far better latency scaling, proposing theoretical ORN designs that are Pareto optimal in latency and throughput.

Motivated by our experiences operating a global WAN, we argue that SDN's reliance on infrastructure external to the data plane has substantially complicated the challenge of maintaining high availability. We propose a new decentralized SDN (dSDN) architecture in which SDN control logic instead runs within routers, eliminating the control plane's reliance on external infrastructure and restoring fate-sharing between control and data planes. We present dSDN as a simpler approach to realizing the benefits of SDN in the WAN. Despite its much simpler design, we show that dSDN is practical from an implementation viewpoint, and outperforms centralized SDN in terms of routing convergence and SLO impact.

In a landmark 1981 paper, Valiant and Brebner gave birth to the study of oblivious routing and, simultaneously, introduced its most powerful and ubiquitous method: Valiant load balancing (VLB). By routing messages through a randomly sampled intermediate node, VLB lengthens routing paths by a factor of two but gains the crucial property of obliviousness: it balances load in a completely decentralized manner, with no global knowledge of the communication pattern. Forty years later, with datacenters handling workloads whose communication pattern varies too rapidly to allow centralized coordination, oblivious routing is as relevant as ever, and VLB continues to take center stage as a widely used — and in some settings, provably optimal — way to balance load in the network obliviously to the traffic demands. However, the ability of the network to rapidly reconfigure its interconnection topology gives rise to new possibilities.

Systems for training massive deep learning models (billions of parameters) today assume and require specialized "hyperclusters": hundreds or thousands of GPUs wired with specialized high-bandwidth interconnects such as NV-Link and Infiniband. Besides being expensive, such dependence on hyperclusters and custom high-speed inter-connects limits the size of such clusters, creating (a) scalability limits on job parallelism; (b) resource fragmentation across hyperclusters.