Traditional edge routing network architecture

Traditional Edge routing networks are networks where the edge is a set of dedicated routers placed centrally in the network. Users usually have multiple Edge networks for services such as business VPN (L2 and L3), subscriber services, internet peering, data center interconnect (DCI), cloud gateways, and so on.

Challenges with traditional edge routing networks

Traditional edge routing networks pose challenges in these areas:

• Distributed services: The bandwidth profile and subscriber density for a distributed edge have changed significantly.

• Failure Impact: The large radius for failures demands more complex geographically redundant homing. It also necessitates the use of network hardware with redundant processors, fabric cards, line cards, power supplies, ports, and chassis. This increases the cost and complexity.

• Service optimization: Service providers might have different edge networks for various services. Optimization of these is difficult. The cost of building a solution that meets all the service requirements on a single platform is high. The most demanding services usually set the price for all others.

• Operational efficiency: Limits operators who want simplicity and efficiency.

• Platform lock-in: Large systems remain in place for many years thereby slowing down innovation.

• Upgrades: Upgrades are delayed as much as possible. Forklift upgrades have become a norm due to compatibility challenges between technology generations.

Business challenges

Internet traffic saw a compounded annual growth rate of 30% or higher over the last ten years. Primary challenges for modern networks include proliferation of devices, increase in end-user bandwidth speed, and the continued movement of applications to the cloud.

Traditional edge routing designs face challenges by new traffic and business realities that include

• ever growing bandwidth with flat average revenue per user (ARPU)

• increasing cost and complexity of scaling centralized network architectures, and

• content and application evolution towards distributed network architectures.

Solution

You must build networks to handle the advanced services and increased traffic associated with modern network services. Networks must evolve so the infrastructure layer can keep up with the service layer. The result of these shifts is driving traffic away from centralized delivery to a more distributed network.

Moving forward, we must consider new network architectures as design options that include these key aspects:

• Fully distributed and fixed routing systems to deliver services at any place in the network

• Fabric models delivering scale-out networks that can be built on-demand

• Flexible deployment options matching provider requirements

• Network simplification at all layers of the network

These are the key benefits of Metro solution:

• Technology benefits:

  • High-capacity edge silicon

  • Convergence of network service functions

  • Flexible network design and systems to fit any size location in the network

• Business benefits:

  • Deliver services closer to users and applications

  • Cost savings

  • Sustainability benefits

• Operational benefits:

  • Improved services resilience

  • Network efficiency

  • Enhanced operations through network automation and orchestration

The Metro network evolution is driven by increasing bandwidth demands, resulting in network functions distributed in the network closer to the end user. This evolution is driving a consequent network architecture evolution. The classical split between access, pre-aggregation, aggregation, and edge leaves room for a more homogeneous network without distinct boundaries between the domains

High-level use cases of Metro solution

The Metro solution architecture covers these high-level use cases:

• Next-generation residential subscriber networks deployments

• Enterprise business services

• Mobile network IP transport

• Centralized and edge data center connectivity including networks that are built to support artificial intelligence

• Internet peering, content delivery, and cloud connectivity

This topic covers the various technologies used in Metro solution.

The Metro solution has the flexibility to support a mix of deployments models based on factors such as service scale, service traffic bandwidth, interface types, and other network constraints.

Metro solution supports these deployment models:

• Standalone MSE: Traditional multi-service Edge (MSE) deployment model for a single Metro edge router where all functions and scale are handled at a single box level. The model is best suited for smaller edge sites which do not need the higher service scale of the Metro Edge Fabric distributed solution. The metro solution enables distribution of devices with MSE functions deeper into the network, increasing overall network efficiency.

• Service on-a-stick: Out-of-line network functions model where the edge services are taken out from the main forwarding path and hosted elsewhere, thereby decoupling the transport and service architectures. The model is ideal for deployments with a high volume of low bandwidth services compared to the volume of transit traffic through the aggregation router.

• Metro Edge Fabric: Scale out Edge services infrastructure model which is a composition of multiple routers in a leaf-spine architecture. The leaf nodes themselves perform the Edge functions of a traditional MSE router but using a distributed horizontally scaled approach as opposed to a vertically scaled approach. There is flexibility in the nodes deployed as a fabric spine, with the primary attribute being they do not terminate many services and are primarily used as transit. The spine nodes could be new nodes in large aggregation sites or re-purpose core nodes already deployed. The key point of the fabric model is horizontal service edge scaling and simplified management using automation.

All models support the Edge as-a-function concept where a service Edge function can be deployed anywhere in the network.

For details on the deployment options for these models, see the Metro Edge Fabric deployment options section.

Apart from the traditional network deployments, the Metro solution architecture covers these new network infrastructure deployments:

• Metro network infrastructure: Modern underlay infrastructure including routing control plane, overlay service signaling, and timing using Precision Time Protocol (PTP), and Synchronous Ethernet (SyncE)

• Edge as-a-function: Edge features at any place in the network (PIN)

• Service on-a-stick: Out-of-line network functions

• Metro Edge Fabric: Scale out Edge services infrastructure

Metro network infrastructure

These are the Metro network infrastructure requirements for all devices participating in the Metro solution.

• Common packet transport: The underlay control and forwarding plane technologies used to support all network services. The common packet transport updates the underlay design outlined in Cisco Converged SDN Transport design. The underlay network is built upon proven technology such as IS-IS for v4 or v6 IGP and Segment Routing using the SRv6 or SR-MPLS data plane.

• Base infrastructure: Additional protocols and technologies outside the scope of routing and forwarding required in the infrastructure to support network services. The solution architecture utilizes PTP G.8275.1 with SyncE for timing distribution, supporting G.8275.2 interoperability wherever required for endpoints requiring G.8275.2.

• Base device automation: Device support for APIs and protocols required to properly automate the network. The solution architecture leverages gNMI-based telemetry using OpenConfig and Cisco native models to monitor all aspects of the network. The interface is used by automation tools such as Crosswork Network Controller and Provider Connectivity Assurance to provide end-to-end infrastructure assurance.

• Base operations: Additional operational tooling and features that are required for operators to properly monitor and troubleshoot the network infrastructure and the services using the network.

• Base security: Base criteria for securing the network infrastructure. This does not cover service security.

• Base QoS: Base set of CoS and QoS feature support to ensure that network service SLAs are met across the network.

Edge as-a-function

The edge as-a-function deployment model removes the traditional location of edge services and allows edge to be served at any point in the network. That is, any router at any place in the network can act as an MSE device that is limited only by the scale required. This is becoming critical and common as higher touch edge services must be distributed to scale. Also, centralizing the edge services is no longer feasible due to service bandwidth growth.

The figure shows the evolution of fabric-based deployment from traditional Edge deployment.

The table lists the characteristics of traditional edge deployment and Fabric-based edge deployment models.

Service on-a-stick

The service on-a-stick deployment is one where service functions are not performed on devices inline between the edge and core of the network. Instead, they are performed on devices out-of-line.

It is inefficient, and cost and power prohibitive to build higher touch network services into the distributed routing hardware when it may not be applicable to all deployments. You can do service on-a-stick deployment either as part of an edge fabric or a more centralized architecture where ingress and egress from the network function are connected to upstream routers. Service scale and bandwidth require a more distributed approach, while compute intensive functions may require them to be located at a centralized data center. The Metro architecture supports both these deployment options.

Service on-a-stick deployment model is suitable for these functions and services:

• High-touch functions that are applicable only to a small subset of traffic

  For example, a NAT appliance capable of handling exception traffic.

• High-touch services that require high performance and scale, such as firewall deployments, where firewalls are placed on a stick to perform security functions

In Metro Release 1.0, the service on-a-stick deployment is applicable to BNG for subscriber services.

Metro Edge Fabric

Creating a network capable of serving a variety of use cases on a common fabric is best served by decomposing a modular chassis into a set of leaf nodes with a specific service role. The disaggregated Edge Fabric functions as a leaf-spine topology with edge service facing ports located on leaf devices connected through fabric ports to an upstream aggregation node. For more details on the Metro Edge Fabric components, see the Metro Edge Fabric section.