Part 1: The Case for Autonomy & Intelligence
It is often noted that service provider networks are more expensive to operate than to procure. OPEX-to-CAPEX ratios on the order of 3-to-1 or more are cited regularly in talks and papers. Optical networks – the foundation of provider network infrastructures – may be among the worst offenders in this regard. Optical networks are, at bottom, high-performance analog systems that manifest complex physics and can be notoriously difficult to plan, configure, optimize, and troubleshoot. This is especially true when they are operated near performance limits to yield the high resource efficiencies that are key to the cost-effective support of steadily escalating capacity demands.
High operational costs have their origins largely in high manual intensity of operational processes. The complexity of optical networks has always driven a requirement for extensive and often highly expert human participation in network operations. This is increasingly problematic as networks continue to scale up and service volumes continue to rise. Further, the “richness” of desired services is also rising, tending toward mass-customization with differentiated and time-varying demands on connection points, bandwidth, and latency. Delivering this richness is a path to increased revenue in enterprise service markets via customized and dynamic premium leased line service offers; it is also a foundational capability to support multiple service frameworks on 5G backbones. But it is extraordinarily difficult to deliver such services with cumbersome operations processes. It is therefore becoming critical to increase the productivity and effectiveness of operations staff and to drive more sophisticated and faster service operations. The key to all this is driving toward increased automation in network operations.
Automation depends on moving operational decision-making from people to software, and empowering software to drive action on the network as a function of decisions taken. The “softwarization” of networks – encompassed broadly by concepts and trends like SDN, NFV, and orchestration – enables in principle the decision-to-action part of this loop. But software-based decision-making depends on software-based intelligence capabilities: in the optical networking regime, “optical intelligence”.
Intelligence for the Optical Network: Toward Optical Network Autonomy
The role of software-based intelligence in network automation is to deliver the information that is needed for sound operational decision-making. Even without closing the automation loop, improved quality – timeliness, accuracy, and so on – of such information improves operational outcomes in terms of achieved capital resource efficiencies and accelerated and improved delivery and performance-assurance of customer services. High-quality “science” brought to software-based network intelligence thus delivers a dual impact: it enables automation while also promoting sustained optimization of operational outcomes. We use the term autonomous to describe a network featuring both qualities. Software-based intelligence is the key enabler of network autonomy: both automation of operational processes and consistently high quality of operational outcomes.
Key optical network operations include resource planning (both green field and incremental); resource optimization (e.g., spectral and related de-fragmentation); service configuration & reconfiguration; service performance & fault management; and service pricing, billing & CRM. Different “key information” – intelligence – is required to optimize these various operations. For example, effective planning and service pricing both depend on the availability of accurate demand forecasting: clearly, more accurate forecasts support both more efficient and timely planning of incremental resource accretions and superior service pricing optimization for more effective revenue attraction. These latter aspects are particularly critical to the pursuit of revenues in the burgeoning premium leased lines market.