Navigating New Networking Trends: Beware of the Urge to Over-Engineer

Scott T. Wilkinson is senior director, technical marketing, MRV Communications. He leads a team of industry and technology experts focused on educating internal resources and external customers on market trends, new technologies, and recent innovations in MRV’s lines of optical and packet product lines.

s_wilkinson_tn

Emerging applications – including on-demand video, cloud services and synchronous replication – are increasing data center interconnect bandwidth demand more than ever before. Operators are grasping at the optical trends they hope will support the mass amounts of data crossing their network. However, it is crucial to recognize that networking approaches that are successful for traditional service providers might not translate into the right approach for the data center.

Service providers and data centers each have a specific role in keeping the data flowing across the network. Service providers offer end-user access and are responsible for transporting user’s information across the network. Data centers are responsible for storing, managing and disseminating that data. Service providers make money from the bits flowing across the network. Data centers make money from the applications and information at the ends of those connections.

Many data centers, however, are actually using equipment and techniques that were designed for traditional service providers. Operators need to be careful they do not fall prey to over-engineering their networks to telecom standards when it’s not required. Instead, they should focus on building a network that delivers the functionality to support current services while providing a future path to meet growing customer demand.

There are many differences between service provider and data center optical networks, but the main areas where data centers can drive efficiency and innovation fall into a few categories.

1. Transport Layer vs. Application Layer
While service providers make their revenue from transport, where reliability and Service Level Agreements (SLAs) drive network design, data center operators make their revenue from applications and content, where cost and capacity drive network design.

Service providers are required to have multiple layers of redundancy in the transport layer in order to ensure the highest level of service and customer satisfaction. Generally, these transport layer requirements do not exist for data centers, as they rely on the packet layer for resiliency.

To increase revenue, data centers aim to quickly boost capacity to locations. Data center operators should focus on innovation at the application layer and consider new options such as leasing capacity and alternate routing. While these alternative approaches would not meet service provider requirements, these options can provide data centers with benefits like a more predictable expenditure model, the flexibility to scale capacity as needed and the ability to better safeguard data by rerouting at the packet layer.

2. Port Density vs. Cost Per Port
For service providers, port density is a paramount concern, while data centers focus on cost per port. This difference is reflected in the split between 100G port deployment by service providers and the continued use of 10G ports by data center operators. Service providers will pay extra for high-density ports, because each port is a revenue-generating entity and translates into more revenue in less space.

Data centers are less concerned with density and more concerned with cost per port and internal network flexibility. With the high cost of 100G ports on routers and the need to re-purpose ports quickly, a network based on dense 10G ports is more appropriate.

However, with the rapidly decreasing cost and size of 100G optics and the ability of emerging pluggable 100G coherent solutions to coexist with 10G signals, data center operators should be preparing for 100G deployments within their network. Deploying platforms that offer plentiful 10G port capacity, and include the option to add on additional 100G interfaces as either 10x10G muxponders or as native 100G interfaces, can be a very effective approach.

3. Fiber Standards: Long-Distance vs. Short-Distance
Long-distance dark fiber is generally difficult to lease and requires more advanced optical knowledge and access to intermediate sites for amplification and regeneration. This usually leaves data center operators considering hardware requirements for shorter reach and metro distances (where fiber is available) and using leased bandwidth for longer spans. Service providers, who own long-distance fiber, must consider hardware requirements for a much wider variety of applications, including long-distance routes that may carry a wide spectrum of signal types.

Because the majority of data centers lease short-distance fiber for a smaller portfolio of service types, there is no need to select networking platforms that are built for long distance, multi-service deployments and can be over-engineered and complex.

4. Support Staff and Simple Solutions
Finally, data center operators need simple, point-to-point solutions that do not have complex operations, are easy to deploy and do not need to be touched after installation. While service providers have a constant support team to organize part numbers, learn complicated software structures, and constantly update software licenses, this is not the case for data centers. Data centers require simple, low-cost solutions that can easily be operated without worrying about the details of optical technology. Optical interconnect should look like virtual fiber, and data center operators should evaluate their equipment choices appropriately.

It can be easy for data center operators to become enchanted by emerging and exciting service provider-designed solutions for optical networking. However, it is critical that the data center operator remains focused on the unique needs—and capabilities—of the data center network. By addressing their current network challenges and opportunities, they can easily avoid over-engineering the network and continue to successfully manage their high profile role in the data revolution.

Industry Perspectives is a content channel at Data Center Knowledge highlighting thought leadership in the data center arena. See our guidelines and submission process for information on participating. View previously published Industry Perspectives in our Knowledge Library.