Abstract:Suitable composable data center networks (DCNs) are essential to support the disaggregation of compute components in highly efficient next generation data centers (DCs). However, designing such composable DCNs can be challenging. A composable DCN that adopts a full mesh backplane between disaggregated compute components within a rack and employs dedicated interfaces on each point-to-point link is wasteful and expensive. In this paper, we propose and describe two (i.e., electrical, and electrical-optical) variants of a network for composable DC (NetCoD). NetCoD adopts a targeted design to reduce the number of transceivers required when a mesh physical backplane is deployed between disaggregated compute components in the same rack. The targeted design leverages optical communication techniques and components to achieve this with minimal or no network performance degradation. We formulate a MILP model to evaluate the performance of both variants of NetCoD in rack-scale composable DCs that implement different forms of disaggregation. The electrical-optical variant of NetCoD achieves similar performance as a reference network while utilizing fewer transceivers per compute node. The targeted adoption of optical technologies by both variants of NetCoD achieves greater (4 - 5 times greater) utilization of available network throughput than the reference network which implements a generic design. Under the various forms of disaggregation considered, both variant of NetCoD achieve near-optimal compute energy efficiency in the composable DC while satisfying both compute and network constraints. This is because marginal concession of optimal compute energy efficiency is often required to achieve overall optimal energy efficiency in composable DCs.

Network-aware compute and memory allocation in optically composable data centers with deep reinforcement learning and graph neural networks

Network Aware Compute and Memory Allocation in Optically Composable Data Centres with Deep Reinforcement Learning and Graph Neural Networks

A Physical-Aware Framework for Memory Network Design Space Exploration

Fast and scalable all-optical network architecture for distributed deep learning

Automatically Reconfigurable Optical Network for HPC System Based on Deep Reinforcement Learning

MONet: heterogeneous Memory over Optical Network for large-scale data center resource disaggregation

Optical Networks for Composable Data Centers

Machine-learning-aided cognitive reconfiguration for flexible-bandwidth HPC and data center networks [Invited]

Accelerate Distributed Deep Learning with a Fast Reconfigurable Optical Network

LOAM: Low-latency Communication, Caching, and Computation Placement in Data-Intensive Computing Networks

Energy Efficient Placement of Workloads in Composable Data Center Networks

Network Topologies for Composable Data Centres

Constrained In-network Computing with Low Congestion in Datacenter Networks

(Com)2net: A Novel Communication and Computation Integrated Network Architecture

Towards an energy-efficient Data Center Network based on deep reinforcement learning

Resource Allocation in Disaggregated Data Centre Systems with Reinforcement Learning

MIND: In-Network Memory Management for Disaggregated Data Centers

Machine Learning Assisted Optical Network Resource Scheduling in Data Center Networks

Scalable Knowledge-Defined Orchestration for Hybrid Optical–electrical Datacenter Networks [Invited]

In-Network Pooling: Contribution-Aware Allocation Optimization for Computing Power Network in B5G/6G Era

Computational Offloading in Semantic-Aware Cloud-Edge-End Collaborative Networks