Many IT administrators are unfamiliar with the differences between data center networking and enterprise networking.
In a recent survey, IT decision-makers from large and small organizations from both public and private sectors on the CDW Customer Advisory Board defined data center networking as the practice of integrating various technical resources located in a data center or server room. Respondents were less sure about the definition of enterprise networking, and one common answer was “end-to-end networking productions and solutions,” a somewhat broad definition. Campus networking was commonly understood to refer to connecting multiple buildings that are part of the same organization and are in close physical proximity, which suggests that respondents see crossover between campus and enterprise networking.
Respondents largely saw both data center and enterprise networking as relying on the same product set: cabling, network infrastructure (such as routers and switches), firewalls and wireless technologies.
While it is true that there is significant overlap in the types of products necessary to build out both data center and enterprise networks, there are also important differences between the two network types. Some of these differences manifest in the specific components that are best suited to each type of network, but traffic patterns, architecture, operations and security are all important differentiators as well.
Robert Herriage, team lead for enterprise networking at CDW, says many organizations still take a one-size-fits-all approach to networking, relying on the same equipment and designs for both data center and enterprise networks. This approach has become less efficient over time, he says, as the rise of virtual machines has driven huge increases in “east-west” traffic between servers in the data center (as opposed to the largely “north-south” client-server traffic that is most common in enterprise networking).
“The biggest difference between the network types is that traffic needs to be optimized differently,” Herriage says. “In enterprise networking, we’re trying to get to the Internet or to the data center. Whereas in the data center, we’re communicating between different servers. The equipment and the features that are prevalent in each type of network are all built around optimizing those different types of traffic patterns.”
The differences between data center and enterprise networking include:
While routers and switches are integral to both data center and enterprise networking, not all hardware is created equal. Some organizations simply deploy switches based on cost or familiarity, but optimized networks will include switches that are designed specifically for one network type or the other. Cisco Systems’ Nexus family of switches, for example, is designed primarily for use in the data center, while the vendor’s Catalyst switches are built for enterprise networking. Similarly, Brocade’s VDX family of switches is best suited for data center networking, while the ICX family is more appropriate for enterprise networking.
“A data center–class switch has much higher bandwidth and lower latency than a typical enterprise switch,” explains Rick Frothingham, a systems engineer at Cisco.
Traffic patterns are the very reason that high bandwidth and low latency are so much more important in data center networking than in enterprise networking. “When you’re running an application clustered over multiple servers, those servers are going to be talking to each other a lot,” Frothingham says. “When you have a data center that runs a web application — a storefront, for example — there are going to be multiple servers responsible for the running of that site, and they’re all going to be communicating with each other.”
“Data center networking is all about more density, more bandwidth,” says Senthil Sankarappan, director of product management for Brocade. This stands in contrast to the more spread-out architecture of enterprise networks. “There, you need to provide a lot more connections, rather than high bandwidth.”
To accommodate this increasing need for bandwidth, data center architecture has moved away from a hierarchical model and toward a “leaf-spine” model in which “spine” switches make up the core of the architecture and “leaf” switches form the access layer that delivers networking connection points for servers. This model allows administrators to scale out the network horizontally, Sankarappan explains. “When the traffic grows, you just add more leaf switches.”
While enterprise network administrators are largely concerned with ensuring that users are able to seamlessly access the Internet and corporate resources, Frothingham says, those responsible for data center networks are more concerned with how quickly they are able to provision new resources. Because of this difference in focus, managers of data center networks are better able to take advantage of the operational benefits of programmable switches, which include automated provisioning and configuration. These capabilities fit in with larger orchestration solutions that provide centralized management of data center resources, helping IT managers to monitor and manage a variety of tools and components from a single location.
“You could get to a point where every enterprise switch is programmable, but what does that get you? Nothing,” Frothingham says. “It doesn’t impact the end user experience at all. Whereas it makes a huge difference in the data center, in terms of managing resources efficiently.”
Another important difference between data center and enterprise network operations is the additional emphasis on power and cooling in the data center.
Because end users access corporate resources via the enterprise network, access management solutions are key to enterprise networking. Many enterprise networks rely heavily on the 802.1X standard, which restricts unauthorized clients from connecting to a local area network through publicly accessible ports.
By contrast, security concerns within data center networks largely revolve around identifying malicious traffic.
Learn more about these important differences by downloading the white paper, "Building the Optimal Network."