Data Centers Move From Metro

Today's data centers are moving out of the metro and into the suburbs in order to cut costs, go green, and ensure geographic diversity. But as data centers begin to move from urban metro locations to facilities that are further away from enterprise end users, the old model of building siloed metro and regional networks begins to break down, giving rise to new challenges.

For example, the shift drives an increase in inter-metro bandwidth requirements, fueling the need for 10 GbE and 100 GbE services. This is exacerbated by the increasing need for on-demand connectivity that comes along with the on-demand cloud consumption model. The relocation of data centers into a regional network footprint also means that connectivity must cross the originating metro network, through a regional network, and back through a terminating metro network to reach the secondary data center.

All of those network handoffs and increased provisioning create huge inefficiencies and a complicated management process. Data center providers should take a number of steps to position themselves, and their networks, to reap the benefits of the remote trend. 

Step 1: Take the best of the data center
Data center networks are primarily packet-based, and because they have been based on Ethernet for decades, the industry has developed design practices yielding very dense 10 GbE products. Smaller products allow data center managers to maximize bandwidth using the least amount of power in a very compact footprint. And trends toward software-defined networking (SDN) and associated applications allow for very agile packet-based networks.

Data center networks, however, are designed under the assumption that equipment is collocated within the same building, or at least nearby. Troubleshooting the network is less complicated than in a geographically dispersed network, because all of the equipment and cabling is within easy reach. Once packet-based traffic is sent outside the building, it becomes part of a metro network, running over optical fiber on telephone poles, under the streets, and buried in underground channels. When something goes wrong with a metro network, troubleshooting becomes far more complex, requiring additional management tools.

Step 2: Leverage the best of the metro network
Metro networks use high-capacity coherent optics and photonics that allow traffic to be transferred across the network at great speed. They also embody design philosophies that allow for very high resilience to ensure connectivity is always available, as expected by end-users.

This availability requires a rich set of optical and packet operations, administration, and maintenance (OAM) tools that allow operators to proactively and reactively troubleshoot their network. This way, when an inevitable failure such as a fiber cut occurs, it can be isolated and repaired quickly, keeping the network operational and maintaining set service level agreements (SLAs).

Step 3: Adopt a converged approach
A converged approach means combining the optical and packet layers into a single network platform and locating IP routers only where needed, such as within the data center. The converged approach is more cost-effective for aggregating and switching lower-speed traffic in the metro network, and far simpler.

We can achieve massive scale with this approach, while maintaining the benefits of Ethernet -- from both a simplicity and cost perspective. Ethernet is ubiquitous because it is relatively inexpensive compared to legacy protocols, and relatively inexpensive because it is ubiquitous (forming a virtuous circle).

Because Ethernet is fast becoming the dominant technology both inside data centers and in the metro networks that interconnect them, it makes sense to deploy networks that are raw and powerful, able to switch and aggregate Ethernet, and reduce the cost and complexity of deploying high-touch functions that are not needed in many parts of the metro network.

Early generations of enterprise data centers (2000s) were built in downtown metropolitan areas to be close to corporate offices when real estate and power costs were manageable, but this model no longer makes sense. By adopting a converged approach, service providers can cost-effectively support the expected traffic growth of multiple 10 GbE and 100 GbE services across geographical boundaries. This approach also optimizes service delivery across a user-to-content domain, differentiating service providers' offerings, and can grow revenue by offering emerging on-demand connectivity services to connect enterprises to the cloud.

To learn more about carrier and cloud Ethernet, attend the Carrier and Service Provider Forum at Interop New York on Oct. 1.  Stan Hubbard, Director of Communications & Research for the Metro Ethernet Forum, will explore how the convergence of Carrier Ethernet 2.0, software defined networking, and virtualization technologies are driving a transformation in service innovation.