When we speak of the world around us today being all ‘connected', most of us are not thinking beyond our gadgets, our software and applications, telephones and the Internet.
But behind these ubiquitous technologies that are constantly transforming our lived experience, there is a grand and sophisticated world of powerful machines — switches, routers, servers and transport channel infrastructure — that power our gadgets, our interactions and keep us hooked on to the network.
The principal communication protocol that has built, sustains and allows our Internet to grow is the Internet Protocol (IP)-based networking system. The hierarchical Internet Protocol suite (with all the related protocols) is responsible for routing the network packets across the Web. And in this IP suite, Ethernet is an important technology that performs ‘switching' of data traffic traversing through these networks.
With the increasing demand of high performance data centres, cloud computing deployments and other heavy Internet services, there is an unprecedented need for robustness and scalability in the services offered by Ethernet technologies.
Ethernet has evolved from the initial 10-Mbps links to a switching speed of 400-Gbps links now being deployed in massive data centres. Though speed-wise, the Ethernet meets requirements, the system suffers from ‘looping', a technological phenomenon that could potentially bring down a network. This has been a concern with Ethernet right from the start.
Ethernet-based networks are labyrinthine for the multiplicity of paths they convolve for the purpose of accomplishing redundancy in links between any two hosts. This maze architecture is necessary to provide alternate paths between network hosts serving as contingency, leading to a substantial amount of redundancy in the network infrastructure and, hence, reducing the efficiency of resource utilisation. The redundancy of passive links are accomplished using the Spanning Tree Protocol (STP).
If not for ‘loop avoidance' protocols such as STP, multiple links would end up causing one of the most dreaded scenarios in any IP-based network — Broadcast storm, which could block all the network traffic causing ‘network meltdown'.
A network cognisant of all its nodes, paths, resources and requirements, which can ‘automagically' configure itself to scale up or down based on the demand, while eradicating the tedious Spanning Tree Protocol and the consequent redundancies would fit the description of Ethernet Fabric.
“Over the decades, Ethernet has evolved as new types of application architectures emerged. Today, data centre networks carry traffic for a diverse set of applications including client/server, Web services, unified communications, virtual machines, and storage — each with different traffic patterns and network service requirements. Ethernet can be used to build shared storage pools, which place stringent demands on the network including lossless packet delivery, deterministic latency and high bandwidth. Combined, these changes drive the next evolution in Ethernet networks: the Ethernet Fabric,” says K.P. Unnikrishnan, APAC Marketing Director of Brocade Communications, which makes the VDX family of Ethernet Fabric products.
When compared to Ethernet with Spanning Tree Protocol, Ethernet Fabric will be flatter, intelligent, resilient, flexible and elastic. These traits are possible because the switches in the Ethernet Fabric will be aware of each other to find out the shortest paths without loops, and only the shortest paths are made active. Ethernet Fabric infrastructure can be integrated with the existing networks.
The Ethernet Fabric industry aspires to eradicate Spanning Tree Protocol in about 10 years.
With the buzz now focussed around cloud computing and virtualisation of IT infrastructure for corporate establishments, the backend of the IP network will have to be ramped up. Ethernet Fabric with its ‘flat' approach would, for sure, be a technology that might forever change the course of networking, taking us along with the networks into a newer realm.