2012 Prediction: Xsigo Systems

Why Fabrics Will Meet the Demands of the Virtualized Data Center in 2012

Data center switching is undergoing a sea change as IT managers grapple with the challenges of connecting the virtualized data center. While virtualization brings opportunities for significant savings, it completely upends the traditional operating model. In the virtualized world, connections are no longer static. Performance requirements can spike with increased VM loading. And capital costs can rise unchecked as IT managers design infrastructures to accommodate growing connectivity needs.

What’s needed is an infrastructure that was designed for virtualization. In other words, we need a switching model that matches the efficiency and flexibility of the virtualized servers themselves. Which is why in 2012, we will see the rise of the data center fabric.

In the past, there was no clear distinction between the switches found in a data center and those used elsewhere in the organization. No distinction was needed because the requirements were similar. Switches simply needed to connect groups of data center resources over a set over static links.

Today however, a cloud data center behaves as a compute fabric where any server can run any application. In this environment, legacy Ethernet switches can result in complex and costly network topologies. Legacy Ethernet routing protocols produce complex data paths that limit bandwidth and add latency. And when something fails, debug can be complex and time consuming.

Data center fabrics address these issues with a new approach to server connectivity. At the highest level, fabric solutions let you manage groups of devices as a whole rather than controlling individual switches. You define the data path end points and let the fabric handle the routing in between. The network itself optimizes performance and resource utilization, thus reducing management complexity.

By decoupling the “connection” from the physical connectivity, a fabric can address the most troublesome challenges of legacy Ethernet.

The “East-West” Problem

It used to be that data centers were laid out in neat, orderly tiers, allowing data to flow in predictable paths. But with virtualization, applications run everywhere. This creates messy server-to-server data paths where data must negotiate multiple layers of infrastructure to reach its destination. The resulting latencies, bottlenecks and routing complexity are called the “East-West Problem.” Data center fabrics will address this with smarter data paths that can traverse the topology much more directly, resulting in better performance and reduced physical complexity.

Exponential Performance Demands

Virtualized servers demand a lot more I/O bandwidth than bare-metal servers do. As result, hard-core virtualization users have hungrily jumped to 10G Ethernet, bare-metal users have transitioned somewhat gradually to higher speed networks. Users find that virtualization works better with more bandwidth. Fabrics enhance performance by eliminating choke points and by enabling faster interconnects.

VLAN Exhaustion

VLANs addressed the problem they were created to solve: LAN segmentation for workgroups. But when it comes to creating large numbers of isolated interconnects within an agile data center, they don’t scale well. They can become a management nightmare. Furthermore, VLANs have limits; you can run out of them. Eliminating this problem is a challenge that must be overcome. A fabric that is designed for virtualization lets you create isolated connections without requiring VLANs.

In 2012, Fabrics will address these challenges with a faster, simpler, more agile infrastructure that will help IT managers virtualize more applications and deliver greater application performance at less cost.