If you're struggling to rearchitect your WAN to handle unprecedented growth, you're not alone. Enterprise organizations are gearing up for a bandwidth tsunami.
According to Nemertes Research's just-released 2010/2011 benchmark study, although a minority (28%) of organizations are predicting an increase in WAN bandwidth, those who are predicting an increase are anticipating a doozy. The average bandwidth increase in 2010 for those who expect growth is 220%. In 2011, they're predicting a whopping 550%.
What's driving the demand for growth? If you're among those experiencing it, you already know: First is the increased dispersion of users to branch offices, coupled with data center consolidation. Simply put, this means more WAN traffic, as IT moves servers farther away from users.
Second is the deployment of bandwidth-hungry applications, particularly videoconferencing and desktop virtualization.
Why does this matter? Because the dramatic increase in WAN bandwidth is driving many organizations to revisit how they're architecting their WANs. These firms are beginning to look at new technologies, including wavelength services and dark fiber. They're also revisiting how they're structuring carrier contracts, investigating regionalized approaches for high-volume portions of the network, and increasingly shifting traffic onto "non-traditional" services, like the Internet and broadband wireless.
The key to an effective next-generation WAN architecture? Match technologies and services to interconnection requirements. Herewith a blueprint:
Data center-to-data center connectivity: Data center interconnectivity requires high-capacity, low latency links. Although multiprotocol label-switching (MPLS) can do the trick, it's often an expensive solution. Telecom architects are increasingly turning to solutions such as carrier Ethernet (either point-to-point or virtual private LAN service, VPLS), or dark fiber or wavelength services. Seventy-one percent of organizations will be deploying some form of carrier Ethernet by the end of 2010, typically for data center-to-data center connectivity.
Why? Three reasons: First is cost. On a per-megabit-per-second basis, carrier Ethernet can run 25% to 50% lower than other technologies. Second is bandwidth: with Carrier Ethernet, users are able to procure up to 10Gbps of bandwidth (the equivalent of 2.5 OC-768 circuits). As a result, as bandwidth requirements increase, carrier Ethernet becomes more appealing. Finally, there's the ease of deployment and management. Users report that carrier Ethernet is straightforward to install, and performs reliably. "Right off the bat, it worked like a charm," says the CIO of a midsized professional services firm.
Data center-to-branch, and branch-to-branch, connectivity: Most organizations (80%) have deployed MPLS, and plan to continue using it for site-to-site connectivity (at least for midsized-to-large sites). Some firms -- mostly leading-edge organizations -- are also looking at rolling out carrier Ethernet for the "core" WAN. The primary challenge? Availability. Carrier Ethernet isn't as widely-available as MPLS.
Remote-site and microbranch connectivity: For smaller sites, companies are exploring a range of connectivity options. An Internet VPN is one approach (and one that's increasingly common). A more innovative approach is to deploy 4G wireless technologies, either direct to the router or to each individual employee. Again, the challenge here is availability: Most carriers are just beginning to roll out broadband wireless (in the U.S. at least). Another problem is capacity: Carriers haven't designed their networks for use as wired-WAN replacements.
The bottom line? When designing your next-generation WAN, think out of the box --- and be sure to match services to traffic flows.
Johnson is president and senior founding partner at Nemertes Research, an independent technology research firm. She can be reached at firstname.lastname@example.org.
Read more about lan and wan in Network World's LAN & WAN section.
This story, "The new WAN architecture: a blueprint" was originally published by NetworkWorld.