Perspectives on the Future of Service Provider Networking: Mass Network Simplification

Traditional service provider networks have become very complex, creating significant overhead across engineering and operations teams tasked with building, expanding, and maintaining them. This results in higher costs, reduced agility, and increased environmental impact. Built on multiple technology layers, domains, protocols, operational silos, and proprietary components that have been stacked over years or decades, service provider networks must go through mass simplification to couple with our society’s increasing business and sustainability demands. Simplification is key to allow service provider networks to continue supporting exponential traffic growth and emerging demands for service agility while reducing the cost of services and power, as well as footprint requirements.

In some sense, talking about why networks need to be simplified is like talking about the importance of exercising for our health and well-being – both can start small and deliver clear, unquestionable long-term benefits, yet we can always find an excuse not to do them. And like many people that struggle to start an exercise routine and maintain it over the long run, many operators struggle to embrace simplicity as a long-term network design principle that benefits the health and well-being of their network.

Intuitively, a leaner network with less moving parts will be simpler, more efficient, consume less resources, and allow for a smoother operation, thus lowering its total cost of ownership. Similarly, using common design, protocols, and tools across the end-to-end network improves agility. Such simplifications can be achieved through small, consistent changes from network design to operations. Over time, networks will achieve compounded benefits in cost savings, lower power use, and improved environmental impact. Operations will be more agile too, directly impacting customer experience.

Mass network simplification is about taking a holistic approach to apply modern network design and operational practices, embracing simplification opportunities across every network domain, and automating everything that can be automated. It’s also about making simplicity part of the engineering and operations culture.

There are several potential areas of simplification to aim for, from the end-to-end network architecture all the way down to the network device level. The following table provides some examples:

Network Level	Mass network simplification opportunity	Examples
End-to-end Architecture	Removing legacy technologies and converging services towards modern IP networks	Moving TDM-based private line and dedicated wavelength services onto IP/MPLS networks using circuit emulation and, thereby, eliminating the need for dedicated legacy SONET/SDN or OTN switching equipment
	Integrating technologies to remove redundancy and lower interconnect costs	Integrating advanced DWDM transponder functions into pluggable optics that go directly into router ports using Digital Coherent Optics (DCO) technology
	Collapsing technology layers, removing functional redundancy, and converging services and network intelligence at the IP/MPLS layer.	Adopting Routed Optical Networking solution which converges L1, L2 and L3 services and advanced network functions, e.g., traffic engineering and network resiliency, at IP/MPLS layer while simplifying the DWDM network requirements as routers are connected hop-by-hop and the IP/MPLS network is self-protected
	Using common technologies end-to-end, avoiding technology and operational silos	End-to-end unified forwarding plane using Segment Routing over IPv6 (SRv6) and an end-to-end unified control plane using M-BGP including EVPN across core, edge, aggregation, access networks and data center fabrics – distributed to the edge or centralized
Device	Adopting modern network platforms with simpler and more efficient hardware architectures	State-of-the-art Network Processor Units (NPUs) based on System on a Chip (SoC) multi-purpose architecture, allowing simpler, more scalable, and more efficient routing platforms
Protocols	Reducing the number of protocols required to run the network	IETF’s Segment Routing and EVPN standard technologies reduces the number of protocols in an IP/MPLS network from 6 or 7 down to 3 (50% reduction) while improving network resiliency and service ability
Management & Automation	Building management and automation solutions based open software frameworks	IETF’s ACTN framework, ONF Transport-SDN framework and OpenConfig gNMI
	Consolidating software interfaces to open APIs and data models	YANG model-driven APIs using NETCONF and/or gNMI , T-API interfaces

Let’s look at two examples of how these network simplifications can be introduced in small steps as part of a long-term initiative – one at the IP/MPLS network protocol level and another at the end-to-end network architecture level.

Mass network simplification in practice

IP/MPLS networks provide unmatched multi-service capabilities. They support Layer 1 services through circuit emulation, Layer 2 services (point-to-point and multipoint, i.e., E-Line, E-LAN, E-Tree services), Layer 3 VPN services as well as various internet services. The technology required to support those services was developed and standardized over many years and as a result traditional IP/MPLS networks require many individual protocols – typically six or seven. Segment routing (SR – IETF RFC 8402 and related) was developed at the Internet Engineering Task Force (IETF) specifically to improve this scenario. By embracing a software defined networking (SDN) framework, segment routing combined with Ethernet VPN (EVPN – IETF RFC 7432) can reduce the number of protocols required in the IP/MPLS network by 50% or more, down to three protocols – segment routing, the interior gateway protocol (IGP) routing protocol, and border gateway protocol (BGP) as a service protocol. Resource reservation protocol (RSVP) and label distribution protocol (LDP) can be eliminated, as can other transport and service signaling protocols.

Segment routing also simplifies network devices because it doesn’t require them to maintain state about traffic engineering tunnels otherwise required by the RSVP-TE protocol. Instead, segment routing embraces an SDN architecture where traffic engineering is supported by network controllers.

Segment routing was created with smooth network migrations in mind. EVPN implementations have also been enhanced to allow for smooth migrations. To achieve that, both allow co-existence of the old and new protocol stack. Co-existence means both traditional IP/MPLS protocols and segment routing are enabled on the same network, or traditional networks can be connected to segment routing networks through routers that provide “interworking” functions to allow traffic to smoothly cross them. Besides that, segment routing was also created with operational simplicity in mind. It’s enabled by using simple configurations since there are less protocols involved. As a result, network operators have been migrating their networks to segment routing for quite some time and have fully transitioned their networks to this much simpler architecture.

At the end-to-end architecture level, service providers also had to stack multiple technology layers. This multi-layer architecture typically has at least four key technology components: IP/MPLS for packet services, OTN switching for TDM grooming and private line services, DWDM transponders for mapping grey signals to DWDM channels, and DWDM ROADMs to cross-connect DWDM channels across multiple fibers. Each technology layer has its own management system and runs its own complex protocol stack. Multiply this for each network domain (WAN, metro, access, etc.) and add a multi-vendor component and the result is a very complex architecture that’s hard to plan, design, deploy, and operate. It’s also very inefficient as it’s hard to optimize all the network resources mobilized for any given service as well as troubleshoot network faults.

Technology innovations made possible the emergence of routed optical networking, a much simpler and cost-effective end-to-end network architecture. These are key improvements promoted by routed optical networking:

◉ Full services convergence at the IP/MPLS layer, including private line services through private line emulation (PLE) technology

◉ Elimination of OTN switching – OTN services are supported by PLE technology

◉ Integration of advanced transponder functions into pluggable optics using digital coherent optics (DCO) technology that goes directly into the router ports

◉ Centralization of network intelligence at the IP/MPLS layer for traffic engineering and network resilience removes the dependency on complex transport control planes (e.g., WSON/SSON)

◉ Use of industry-defined open interfaces and data models for management and automation with segment routing to further simplify the end-to-end network

Even such a breakthrough network transformation like routed optical networking can start small. The first step can involve simply replacing transponders with DCO pluggable optics as you adopt 400GE in your IP/MPLS network, while maintaining your existing DWDM network. In parallel you can start your path towards segment routing adoption. Over time, you can embrace more automation and start migrating TDM services to the IP/MPLS layer, until you eventually adopt all the innovations and deploy a full featured routed optical network. As we speak, many service providers have already started these network transitions.

Source: cisco.com