Growing traffic demands are putting more pressure on campus core Ethernet switches. Here in the first part of this series, we examine the steps needed to plan for a core switch upgrade. The second part of the series looks at howJuniper, Cisco and HP differentiate their switches.
Campus core switches sit at the heart of campus networks and are responsible for high-speed routing and switching. Traffic growth at the edge of the campus network will affect the core, so it's essential for campus core switches to be able to respond to heavy influxes of traffic and continue to keep packets moving swiftly.
In enterprises, IT consumerization and mobility are pushing many older campus core switches to their limits. Universities have dealt with consumerization for years, but even they need to upgrade campus core switches. And they're focusing on options for core 100 Gigabit Ethernet, since many are deploying desktop virtualization. Running 2,000 computer lab seats from central VDI servers can consume 7 to 10 Gbps worth of bandwidth if booted simultaneously. Researchers are also working on projects that require transferring petabyte data sets.
Campus core refresh: Long-term planning
"Forklift overhauls" or refreshes of campus core switches typically occur every five to 10 years, or longer, depending on the switch. Because these switches are the foundation of your network, it's crucial to factor in future requirements and choose an architecture that can continue to support your needs.
"For our environment, it's at least 10 years for a forklift refresh," said Joe Rogers, senior network engineer at the University of Southern Florida (USF).
The modular nature of most campus core chassis allows network engineers to make incremental upgrades to the core over time. "For example, we might upgrade supervisor engines or fabric cards as newer-generation hardware becomes available -- every five to seven years," Rogers said.
Factors to keep in mind when refreshing campus core switches
How should network engineers compare the different options for a campus core switch? "Cost, scale and operation," said Eric Hanselman, research director of 451 Research's networking practice.
One of the biggest differentiators in campus switches is the cost per port in density configurations and across switch families within a single vendor's offerings. Cost is simple, but must be traded with capacity, Hanselman explained. "Scale is the density of ports at the necessary speeds. Do you want a single chassis or will a stack do as well?"
Engineering teams should approach the operational consideration from two perspectives: "Both in terms of personnel skills, such as how long it takes to get operational mastery of new gear, and what operational constraints do the systems present? Do you have to take the whole stack/chassis down to do a software upgrade?" he said.
The ability to perform in-service software upgrades (ISSU) can minimize downtime -- critical for these switches. Upgrades are also a real consideration for any new switch installation and can be one of the limiting factors in uptime for an installation, according to Hanselman.
"The operational characteristics of any switch can be as important as the initial cost," Hanselman added. "These are both driven by the management functionality and service requirements. Configuration flexibility is driven by the management software offered with the switch."
What key features do network engineers look for?
As network engineers shop for a new campus core, they examine each switch's price, supported bandwidth, port density, features and manageability.
While the price of a campus core switch is always important, its value is considered even more important. "We're vendor agnostic, so the operating system and platform are irrelevant as long as the platform meets our requirements and is maintainable by our standards-based tools," said Nick Buraglio, a network engineer at a large university.
Bandwidth and density
Bandwidth and 100 GbE port densities were "the most important factors," said Rogers, for the campus core upgrade USF has underway. But at the same time, he cited features such as being able to support software-defined networking (SDN) "as critical" to meeting any future research needs that involve big data.
The networking team at Brandeis University is looking for core/distribution switches with support of standard-based routing -- like Open Shortest Path First (OSPF) -- and chassis virtualization or stacking, such as Cisco's Virtual Switching System (VSS) or Juniper Networks' Virtual Chassis technology, according to John Turner, director for networks and systems at Brandeis.
"Standards and compliance are first and foremost," Buraglio said. "We're also interested in MPLS, VRF-Lite and complete IPv6 support."
Given the long-term investment of campus core refreshes, engineers also want to see a roadmap toward SDN and OpenFlow.
For small network engineering groups, the variation of network operating systems from one vendor to the next can be a purchasing factor.
"With five network engineers managing roughly 80,000 switch ports along with the core infrastructure, it's important to limit the number of different vendors' gear we need to support," said Rogers. "There are differences -- even between different platforms from the same vendor -- so we look at new equipment from these companies to make sure we understand the operational impact of having to manage different operating systems."