People say that in travel, getting there is half the fun. But in network migration, getting there is often almost the whole problem.
It may be interesting to think about the next-generation network as a broad and rapid revolution. Revolution may seem exciting, but the enormous value of existing network infrastructure and the enormous risk of massive technology change make a classic forklift upgrade impractical. We're almost certainly going to advance into the future in smaller steps. The question is how those steps will be taken. Let's take a look at the main options in terms of how to structure your next-generation network migration: the network layer approach, the geographic approach and the service-oriented approach.
The layered network migration approach
From a technology perspective, the most logical way to modernize network infrastructure and operations is in layers. The venerable OSI model says a service network consists of a physical Layer 1 (optics); Layer 2 data-link services, which include Carrier Ethernet; and Layer 3, a service layer that is most often IP. Operators usually think of networks in terms of the transport layer and connection layer.
If you're modernizing, one question is whether you should start at the top or bottom of the stack.
Optical and data-link equipment have a single mission: carry traffic reliably. It isn't rendered obsolete by changes in market requirements, so it typically has a much longer installed life. On the other hand, in many networks (including mobile infrastructure) optical transport is being added regularly to increase transport capacity. Each additional deployment is a point where technology changes could be considered. Since the lower layers don't actually present services to the user, differences in technology there wouldn't impact the services themselves.
The lower layers also make up a fairly small portion of operations expense for network operators because they focus on aggregated traffic, not on connections and services. Operations processes actually cost operators more than capital equipment, and most network operations costs relate to services, which happen at the top layers of the OSI stack. So, if next-generation networks were driven by a need to reduce Opex, it would make sense to start at the top where most of the exposure lies.
But does a by-layer migration model even make sense? Sometimes yes, sometimes no. Both bottom-up and top-down migrations would be the optimum approach in some situations, but there are no clear answers here.
The geographic network migration approach
It is also possible to think of next-generation network migration by geography. The consumption of services isn't uniform across all markets, and even within a given country or market area, there are vast differences from place to place. Areas with high demand earn more revenue for operators and generate more need to change or augment technology. Since there's more infrastructure to evolve in high-demand areas, introducing a new technology would change a greater percentage of total Capex and Opex.
Management centers tend to be geographically divided, as are operator support workforces. Since it's difficult to ask operations people to work with different management/support systems and difficult to troubleshoot in such situations, it would make sense to target a geographic area for modernization and change both equipment and operations practices at the same time.
The problem with this approach, however, is that business services in particular cut across many geographies. That poses the risk of having a customer with some of their sites served with legacy technology and others with next-generation network technology. The lack of homogeneous infrastructure means that customers might not see the same service properties across all sites or be able to use common customer management practices. So, for multisite business services at the very least, the geographical approach has challenges.
The service-based network migration approach
The final option is to consider a migration to next-gen networks by service. We've already seen that network functions virtualization (NFV) is deploying first in managed VPN/VLAN service applications, and that the early applications of SDN are in cloud computing. Timing the introduction of a new service type, like managed services, with a migration to next-gen technology is a particularly good idea because there are no previous customer experiences or operations practices to consider.
Another area where a by-service model of next-gen migration has gained acceptance is in mobile services. Mobile services are the most profitable services to network operators, and mobile traffic has the fastest growth rate. These combine to make mobile infrastructure the focus of more and more operator Capex, which means that there's new opportunity to introduce next-gen technology generated every year.
Most mobile infrastructure technology changes focus on the IP Multimedia Subsystem and Evolved Packet Core elements, neither of which is directly visible to the customer. Both have well-defined interfaces and any technology that can meet the specifications can be interchanged with any other.
The problem with service-based transformation to next-gen infrastructure and technology is the scope of impact. Unless a service has a very contained customer base, changing the technology could impact the full scope of infrastructure and operations. That would either force operators back to the forklift upgrade they want to avoid, or confront support personnel with two different infrastructure models throughout the evolution.
Making practical next-gen network migration decisions
So, what's the solution? For most operators, it's "all three somewhere and none of the three everywhere." The service-based approach has gained the most acceptance, particularly for managed business services and mobile infrastructure. With managed services, any customer-visible changes are covered by the fact that the customer is explicitly changing services, and with mobile infrastructure, the changes are largely at lower network layers not directly visible in the services themselves.
The real lesson, though, is twofold. First, there are no strategies or even combinations of strategies for infrastructure evolution that quickly move networking to the new era. Second, the diversity of approaches to next-generation network migration virtually ensures that pieces of networks and services will evolve under their own local cost/benefit drivers. Somehow, they'll all have to lead to harmonious capital and operations practices, or the problem of creating inefficient silos will emerge and slow progress for everyone.
So good together: NFV and SDN drive change to create flexible network connections
A look at NFV and SDN as providers' mainstream options
Follow the vendor road map: Reality check for potential SDN adopters