This vendor-written tech primer has been edited by Network World to eliminate product promotion, but readers should note it will likely favor the submitter's approach.
The advantages of moving data centers to remote locations are clear: space is lower-cost, power is cheaper and geographic diversity is built in. Google, Amazon and others, for example, have located large data centers in rural Washington State close to hydroelectric dams that deliver low-cost green power. But these locations are up to 300km away from the closest regional points of presence in Seattle or Portland, which challenges the old model of building siloed metro and regional networks.
Enterprises need to evolve from traditional metro-regional/long-haul net designs toward a user-to-content model where the programmable network can dynamically allocate resources, connecting users to the content and bandwidth they want whenever they want it.
By blurring the boundaries of metro-regional networks using a user-to-content approach, enterprises can cost-effectively support the expected growth in 10G and 100G services across geographical metro-regional boundaries at distances of hundreds of kilometers.
This approach also optimizes content delivery across a user-to-content domain, and can apply whether you choose to "Build your Own" network or lease infrastructure through a service provider's managed service.
This decision to build a private optical network versus lease as a service is not just as simple as favoring CAPEX to OPEX or vice versa although that comes into play. Some of the questions that need to be answered before an enterprise makes the leap include:
* Do you anticipate significant WAN bandwidth growth requiring multiple 10G wavelengths? Many large data centers need to be interconnected across metro-regional boundaries at very high capacity, say multiple 10G, 40G, or even 100G services. With a traditional metro deployment model, providing this high capacity is expensive and slow to implement.
The alternative is to directly connect the data centers across metro-regional boundaries, creating an express wavelength, which blurs metro-regional boundaries at the optical layer. This approach consists of extending the regional high-capacity wavelength (40G or 100G) directly to the enterprise data center using regional (1,000 km) reach transponders at the enterprise data center and a passive bridge in the service provider's regional PoP. The cost of leasing dark fiber across 100s/1000s of km needs to be balanced against the cost of managed N x 10G/100G wavelength services across these distances.
* Does your network need to reach across metro-regional boundaries? Not all data center sites require the high-static demand of multiple 10G links described in the previous example. For lower-demand sites, it is not cost-effective to design express wavelengths for each individual datacenter. A more cost-effective approach is for traffic from multiple low-demand static sites to be aggregated into a single, high-capacity 40G/100G link into the enterprise regional backbone.
In this scenario, where you are using the traditional approach of building metro networks with strict metro-regional boundaries, the number of back-to-back metro DWDM terminals increases with the number of manual connections from the metro to the regional backbone.
For more efficient delivery of traffic across metro-regional boundaries, an enterprise can aggregate traffic at the Optical Transport Network (OTN) layer at the enterprise regional datacenter. In this scenario, the low-cost metro DWDM customer-located terminals are left intact at the enterprise data centers, while the back-to-back metro DWDM terminals connected to the regional transponder are collapsed into a single OTN packet/switch node with metro short-range optics facing the enterprise data centers and high-speed optics facing the enterprise regional backbone.
This approach leads to cost savings from increased network velocity and creates a foundation for introducing bandwidth on-demand services across the organization.
* Do you have periodic need for 'on-demand' bandwidth? Consider the regular data backup you do over the weekend. In this example, you need to procure a 10G service for three hours every Saturday night to maximize application performance. With static connections, enterprises would usually deploy service capacity that is typically required for day-to-day operations, such as a 1G service. As a result, a backup operation using static connections would require a much longer duration.
Another application that might include a spike in demand is distribution of media content. You many need, for example, 10G services to distribute high-definition pre-production, uncompressed video streams from event venues to production studios.
On-demand services can be delivered on OTN/packet switching network architectures by adding Software-Defined Networking (SDN)-type capabilities. A programmable and switching-capable network enables multiple enterprise users to utilize the same network resources at different time intervals.
Ultimately, there is no black or white answer on the decision to build or buy it's often many shades of grey. But as you move data centers from metro core to regional locations, consider the questions and evolution path you want follow to optimize the cost of delivering new services and matching service bandwidth to the demand bandwidth in a more accurate way.
Ciena (NYSE: CIEN) leverages its deep expertise in packet and optical networking and distributed software automation to deliver solutions in alignment with its OPn architecture for next-generation networks. We enable a high-scale, programmable infrastructure that can be controlled and adapted by network-level applications, and provide open interfaces to coordinate computing, storage and network resources in a unified, virtualized environment.