The modern data center is changing, fast.
As data center trends continue to predict 25-35% annual growth traffic, data center speed is now everything. Today, this means data center managers are facing the challenge of future proofing, to be able to provide the best, most flexible course of action in the face of growing capacity demands.
However, demand for network capacity can only be supported by a shift to higher switching speeds and, as a result, migration strategies continue to evolve; with the growing affordability of 100G switch links – multimode and singlemode – many companies are now able to update their switch networks from 10G to 100G. We’re also seeing that a shift to a 25G lane is increasingly commonplace, and that the entrance of proprietary and standards-based PAM-4 modulation has brought in the introduction of 50G lane rates.
Looking to the future, this evolution and demand is set to continue:
Ultimately, the challenge in network design today is how to accommodate massive and growing amounts of traffic both now and in the future.
A traditional data center architecture uses a three-layer topology.
Although this design provides a predictable foundation for a scalable data center network, it is less ideal when it comes to supporting today’s low-latency, virtualized applications. For this reason, data center managers are increasingly opting for a “leaf-and-spine” architecture which is optimized to move data in an east-west flow, enabling servers to collaborate in delivering cloud-based applications.
Networks are spread across multiple lead-and-spine switches, making the leaf-and-spine switch layer critical for delivering maximum scale and performance. Each leaf switch is connected to every spine switch which creates a highly resilient any-to-any structure, and the mesh of fiber links creates a high-capacity network resource or “fabric” that is shared with all attached devices. As all fabric connections run at the same speed, the higher the speed the higher the capacity of the mesh network or “fabric network.”
With the increasing density of networks, cabling connectivity and management become more important. Fabric networks require high-speed links across the entire mesh, which often spans the entire data center. It is becoming increasingly common to deploy more links with higher speeds and longer reach.
Applications standard organizations, especially the IEEE 802.3 (Ethernet) and ANSI/T11 (Fibre Channel Committees) have been updating the recommended guidelines to keep pace with rapid increases in bandwidth. These standards aim not just to facilitate the evolution of ever-increasing line rates but they also encourage the development of higher-speed applications, which will increase the cost-effectiveness of links between the data center equipment. A number of intermediate speeds are being developed to fill the gap between 10G, 40G, 100G and 4000G – with standards for 50m, 100G and 200G, 200-Gigabit Ethernet, and 400-Gigabit Ethernet still in process.
Migration to higher line rates offers a solution to the challenges above.
Today’s network design has to reflect massive amounts of traffic, and crucially enable server, storage and network capacity to be scaled up independently – with as little disruption as possible.
To futureproof data centers and see continued success, data center professionals must support higher server densities, deploy more fiber and accelerate their plans to migrate to higher speeds in their core and aggregation networks.
Find out more about high speed migration in the data center here.