This blog is part three of a four-part series discussing Cisco’s routing and automation strategy, the trends and inflections we see in those areas, and how we guide our customers through their digital transformation journeys. See the introduction blog here.
In the first blog of this series, Kevin provided a retrospective view of the trends and waves of innovation we have observed for the better part of the last decade – centered-on traffic localization, transition to all-IP with video content as the primary contributor, and how the center of gravity in networking is moving closer to the end users.
The most recent wave of inflection points is now redefining the overall metro landscape:
- Edge computing and the hosting of virtualized network functions, new revenue-creating applications, and localized content drive the intense need for efficient integration into networking.
- Rapid evolution of connectivity ecosystems, partnerships, and value chains. The edge and middle-mile spaces are becoming ever more crowded from disruptive entrants without a legacy of complexity in infrastructure, strategy, and OSS/BSS.
- Growth in Broadband access is fueled by hybrid work, insatiable demand for content, device proliferation, and large investments in fiber access to close the digital divide. With the adoption of SD-WAN and SASE, the Internet is becoming the backbone of everything.
The routing layer has become the undisputed focal point for secure connectivity. What used to be separate infrastructure for business, mobile, and residential networks have become one. The protocol stack has been reduced to revolve around BGP VPNs and Segment Routing. Thanks to the routing silicon evolution and the emergence of interoperable, pluggable, coherent optics, IP and Optical networks are converging with the metro as the natural starting point. These network realizations have been the foundation of Cisco’s Converged SDN Transport strategy.
Today, Acacia’s recently launched high transmit power Bright 400ZR+ extends the same deployment and operational simplifications of Routed Optical Networking (RON) to multi-span and ROADM-based applications. The RON architecture highlights the generic benefits of router-hosted pluggable functions in sparing, reusability, power, space, and per-module pay-as-you-grow on a converged infrastructure.
The distributed hosting of virtual network functions and content drives the need for seamless and secure integration of networking and the cloud, with the metro as the focal point. This collapses and extends the requirements of the traditional roles of access, aggregation, and edge. From hierarchical points of simple multiplexing, aggregation, and backhauling to also:
- Serve as a programmable fabric exposing its capabilities and SLA characteristics which any overlay service can leverage for monetizable differentiated services.
- Seamlessly provide service anchors and secure cloud on-ramps anywhere and uniformly as requirements for hosting and peering change over time.
- Support on-premises, hybrid, and public cloud environments with a well-behaving IPv6 routing stack as the lowest common denominator. The next and final blog in this series will cover this topic and its 5G context.
Unified Metro Fabric
Cisco Silicon One unifies four key segments across multiple devices: the hardware architecture, software APIs, P4 forwarding code, and flexible run-to-completion engines. This unification of function with the ability to right-size scale and bandwidth to meet current and future real-world requirements enables consistent capabilities across the portfolio – without the overhead in hardware and software typically associated with such flexibility.
When mapping silicon-level considerations to system-level benefits, a pattern of simplicity emerges. As routing silicon bandwidth will continue to outgrow that of most network roles, the preference and viability for single-chip systems in the metro fabric is evident due to the:
- Hardware and software complexity associated with scheduled multi-chip fabrics.
- Each additional component consumes space and power, adds cost to the overall solution, and comes with its own probability of causing (partial) failures.
- Direct path to the alignment of spend-to-service generated revenue.
When working with our Service Provider customers, we find that many could serve entire metro regions with a couple of P100-based single-chip spines, each yielding 19.2 Tbps. If built in a compact (fixed or centralized) form factor with a single piece of silicon in the forwarding path; cost structures, latency, operational simplicity, reusability, space, and power efficiency would all favor that model. In Figure 1, the leftmost picture represents a solution built on a single P100. The other two represent what it would take for Cisco or anyone else to achieve anything comparable with other routing silicon.
Figure 1 – Single-chip system based on P100 versus alternative solutions
That shift to, and proliferation of, compact system form factors is nothing new. This inflection point was one of the main drivers behind the launch of the Network Convergence System (NCS) 5500/5700 portfolio in 2016, which also captured the first wave of a more aggregating transport. In addition to the significant trend for conventional fixed systems in stand-alone or Clos fabric topologies, a reasonable compromise can be achieved with centralized systems. They provide cost structures and power efficiency closer to a fixed system but with the flexibility of a modular system in terms of redundancy options, upgradability, higher radix, and interface mixing.
Modular systems, in contrast, bring many considerations around the lowest common denominator chip variants, accounting, failure scenarios, and the need to fully populate the system to realize value fully. However, the idea of investment protection in a modular system does play out in highly dense and static areas of the network. A combination of single-chip, fixed, and centralized systems, can be used in the metro. Where needed, horizontal growth augmentations can be realized without the conventional cross-functional dependencies. Where built and capacity planned for 1+n redundancy with fiber as the lowest common denominator, Service Providers are starting to question yesterday’s truths in network architecture. As a result, the overall service availability focus shifts from system-centric to network-centric, allowing for flexible scaling models, simpler and efficient building blocks, and repeatable designs as the foundation of a unified metro fabric.
Irrespective of underlying silicon, infrastructure, and topology, IOS XR is the underpinning element and anchor for operational automation and convergence. For metro deployments, creating operational ease in mixed security environments is a must, and IOS XR delivers through:
- Simple: Reduced footprint, support for on-box Linux applications, and kernel programming. Customers can manage the entire life cycle of XR RPMs, Linux RPMs, and Golden ISOs. License pools are shared network-wide
- Secure: Hardware-anchored root of trust, secure-boot process, control and data plane encryption, Integrity Measurement of all processes at runtime, signature validation of all installed RPMs, secure zero-touch provisioning
- Modern: Model-Driven APIs for configuration and operational state, Service Layer APIs, open programmability, and industry-leading implementations of BGP VPNs and Segment Routing
- Versatile: Runs on Cisco-powered custom and merchant silicon, selected OCP-compliant hardware, and x86 data plane
We also recognize that most networks deployed are anything but simple, immutable, or heterogeneous after potentially decades of accumulated complexity. In addition, the foreseen increase in the number of policy points and devices require operational models and tooling built to natively support such environments throughout the life cycle. Cisco Crosswork provides complete and normalized service-to-fiber context and end-to-end automated operations across vendors, technologies, layers, domains, and services.
Unified Edge Hosting
The network edge must become a floating function implementable at any hierarchical layer to facilitate integration with more distributed computing and content hosting. However, it cannot come at the cost of the traditional hardware and software overhead associated with such flexibility. Instead, through a converged services architecture and innovations in routing silicon, capabilities are consistent in behavior, and the distribution level becomes the primary proxy to bandwidth and logical scaling requirements. This leads to the complete collapse of the traditional metro roles, as visualized in Figure 2.
Figure 2 – A multi-dimensionally converged metro edge hosting architecture
The other complementary principle is decoupling transport and service architectures. With the same considerations as any other pooling of resources, the hosting occurs outside the main forwarding path and uniformly across high-touch routing silicon and hybrid cloud infrastructure. This enables the promise of isolated complexity and life-cycle management, resource optimization, and “any service, anywhere” concerning universal access leaves and ports.
Recent Cisco innovations in Segment Routing v6 and micro-SID provide the glue to advanced programmability and end-to-end traffic engineering, chaining, assurance, and security isolation across the fabric. Along with rich observability, the foundation of a consumable Network-as-a-Service construct that spans all the traditional policy insertion points is formed. Any overlay service, whether traditional VPN or Cisco SD-WAN with ThousandEyes WAN Insights, would subscribe to the transport fabric’s services and the associated performance SLAs. A well-behaving IPv6 routing stack becomes the lowest denominator for extensions into other domains.
With the trends mentioned above in broadband Internet growth and proliferation of content, delivery economics and value chains come into focus. At the same time, numerous inflection points have surfaced that call for an evolved subscriber edge architecture. Broadband Forum-driven standardization of control and user plane separation (CUPS), wireless/wireline convergence, and dynamic session steering bring disruption and simplicity to an otherwise closed ecosystem with vertical dependencies and lock-in. This is at a time when it must be recognized that subscriber profiles and the associated queuing requirements have evolved along with the convergence of access technologies and policy layer. To efficiently operate and scale a growing distributed floating network edge model, Cisco’s subscriber edge solution is built for a standardized disaggregated environment that enables independent control/user plane economics, fewer OSS touchpoints, higher feature velocity, optimized IP resource management, and rich visibility and automation.
According to Cisco’s latest VNI, at least 70% of all consumer Internet traffic is video by the end of 2022. Delivered over centralized peering points from CDNs, either directly and/or via per publisher caching infrastructure, Service Providers are rarely able to monetize that traffic to broadband subscribers.
Through these innovations in silicon, systems, optics, software, and architectures, we continue to redefine the economics of the Internet. With metros as the center of gravity for network and cloud convergence combined with the proliferation of broadband Internet and content, it becomes the obvious starting point for optimizing delivery economics. At this intersection, Service Providers can also begin to monetize their edge and their unique position as the natural metro partner and multi-cloud glue in rapidly evolving edge and middle-mile value chains.
Please stay tuned for the next blog in this series, focused on Converged SDN Transport for 5G – Cisco’s cloud-ready solution.