A Scalable Multi-Stage Packet-Switch for Data Center Networks

TL;DR

This paper introduces a scalable, efficient multi-stage packet switch for data center networks using Networks-on-Chip technology, improving throughput, scalability, and cost-effectiveness over traditional architectures.

Contribution

The paper proposes a novel Clos-UDN switch architecture that replaces complex modules with NoC-based central stages, enhancing performance and simplicity in data center network switches.

Findings

Outperforms traditional multistage switches in simulations

Provides high throughput and load balancing

Reduces complexity and cost of switch design

Abstract

The growing trends of data centers over last decades including social networking, cloud-based applications and storage technologies enabled many advances to take place in the networking area. Recent changes imply continuous demand for bandwidth to manage the large amount of packetized traffic. Cluster switches and routers make the switching fabric in a Data Center Network (DCN) environment and provide interconnectivity between elements of the same DC and inter DCs. To handle the constantly variable loads, switches need deliver outstanding throughput along with resiliency and scalability for DCN requirements. Conventional DCN switches adopt crossbars or/and blocks of memories mounted in a multistage fashion (commonly 2-Tiers or 3-Tiers). However, current multistage switches, with their space-memory variants, are either too complex to implement, have poor performance, or not cost effective. We propose a novel and highly scalable multistage switch based on Networks-on-Chip (NoC) fabrics for DCNs. In particular, we describe a three-stage Clos packet-switch with a Round Robin packets dispatching scheme where each central stage module is based on a Unidirectional NoC (UDN), instead of the conventional single-hop crossbar. The design, referred to as Clos-UDN, overcomes shortcomings of traditional multistage architectures as it (i) Obviates the need for a complex and costly input modules, by means of few, yet simple, input FIFO queues. (ii) Avoids the need for a complex and synchronized scheduling process over a high number of input-output modules and/or port pairs. (iii) Provides speedup, load balancing and path-diversity thanks to a dynamic dispatching scheme as well as the NoC based fabric nature. Simulations show that the Clos-UDN outperforms some common multistage switches under a range of input traffics, making it highly appealing for ultra-high capacity DC networks.