Programmable carrier networks, a new, eclectic, emerging architectural approach, incorporate concepts such as SDN, NFV, shared mesh protection, path computation element protocol and cloud computing concepts and blends them with established transport, switching, routing and network management techniques. The objective of this merger is to overcome the barriers created by traditional network architectures that carriers are encountering as they try to accommodate high and volatile traffic volumes and unpredictable traffic patterns as well as respond to the innovative business models of cloud-based and OTT service providers.
Traditional network architectures employ many purpose-built hardware devices and multiple proprietary element management systems. Complex systems integration projects are required to add capacity and introduce new services. Despite the system integration efforts many manual interfaces are retained. This causes slow service delivery and limits subscribers' service options. Other consequences of this difficult to manage and inflexible architecture include very poor network capacity utilization and the failure to fully exploit the potential of mesh network topologies. The result is very costly network infrastructure and the need to lock subscribers into multiyear contracts to cover high fixed costs. This architectural approach is destroying carriers' competitiveness, especially as more nimble cloud-based and OTT service providers enter the market.
Programmable networks use cases recently published by ADVA, Brocade, BTI, Ciena, Cisco, Cyan, Huawei, Infinera and Juniper describe solutions to overcome these barriers.
Programmable traffic engineering uses SDN principles such as centralized controllers, northbound APIs to traffic engineering software and southbound APIs to network elements to discover and implement optimal traffic engineering solutions within hours rather than months. The traffic engineering software provides visualization, analysis and optimization tools that design, plan and operate the networks. In two recent studies I calculated TCO savings of 35 to 50 percent produced by greatly improved capacity utilization compared to the present mode of operations.
Real-time network self-optimization, an extension of the programmable traffic engineering use case, is an autonomic networking solution that optimizes network traffic for short time intervals. This solution scavenges bandwidth and redeploys it to increase network capacity utilization. A recent study showed a 25 percent TCO savings compared to manual traffic engineering methods.
Multilayer optimization use cases provide globally optimum link capacity solutions across the transport and packet layers. The solutions are implemented automatically using an SDN centralized controller and northbound and southbound APIs. Today, many network operators manage the transport and packet networks through separate operating units in their organizations. This causes lengthy capacity planning and provisioning cycles and substantial capacity over provisioning. This use case compresses the planning and provisioning cycle by integrating the planning process and automating provisioning. The same approach can be used to optimize capacity allocations across multiple aggregation and access network technologies, such as optical, microwave, wireless and DSL. Cost reductions can be as high as 75 percent.
Shared mesh protection schemes also can be automated using the programmable network architecture because they simultaneously deliver higher reliability, multiple failure survivability and lower cost by tapping into the full potential of mesh networks. The programmable network architecture can handle multiple complex failure scenarios in less than 50 ms, automate northbound interfaces to calculate protection paths and interact with planning tools. When compared to 1+1 protection the shared mesh protection use case has a five-year 27 percent lower TCO.
The ability of programmable networks to identify unused bandwidth and redeploy it in near real time also creates revenue opportunities that are not feasible when service offerings are tied to multiyear contracts. Two examples are bandwidth calendaring and bandwidth on demand. Bandwidth calendaring is a prescheduled bulk data transfer service with guaranteed bandwidth and quality of experience. It is billed on a usage basis, which makes this occasional use service much more affordable than if it were priced as permanently nailed-up bandwidth. Cloud backup and disaster recovery are two possible applications of the service. Bandwidth on demand, a similar service offering, provides an enterprise an affordable alternative to incurring network performance degradation during peak usage periods. Workload migration with VMware vMotion is one possible application. These services create new revenues with one-year ROI that ranges from 55 to 94 percent. The returns are high because the required bandwidth is taken from the pool of unused bandwidth that exists under the present mode of operations.
The use cases illustrate cost savings and new revenue opportunities that are created by the ability of programmable networks to provide a closed loop of information flows that continuously update high-level business processes such as traffic engineering on the state of all network elements while sending automated network control command down to multiple vendors' network elements and through all network layers.
Unlike data center networks, OpenFlow is only one of many southbound APIs used in carrier networks. Others include TL1, Corbal, Netconf, SNMP and CLI. This mix of APIs is necessary because of the greater scale and complex topologies of carrier networks, the need for interworking with many service providers and the very large embedded equipment base of the telecom industry. Also, the original SDN efforts were directed to the packet layer while much of the economic promise of programmable networks requires multilayer solutions. Integrating many varying APIs has spurred the practical implementation of programmable network solutions.