The case for disaggregated routing

30 September 2020 | Tim Doiron


Tim Doiron

Blog Author | Guest contributor


From improving variety to network velocity, and economics, Tim Doiron, Infinera’s sr. director, solution marketing, makes the case for disaggregated routing

Dubbed 10G, cable and multiple-systems operators (MSOs) continue to evolve their hybrid fiber coax (HFC) networks to support a path to 10 Gb/s broadband service delivery and the launch of new services including 5G. 

Like many network elements, routers have been chassis-based, with centralised switching modules serving multiple line cards.  Low-speed traffic is aggregated at the network edge into higher speeds and transported toward the core.  The operating system software, including the protocol stack (IS-IS, BGP, OSPF, VRRP), and chassis hardware are tightly integrated, sold and supported as a unified solution from a single vendor.  While this architecture has served us well to date, there are three main limitations to this approach – limited variety, reduced networking velocity, and clunky economics.

Variety and More Choice

Disaggregated routing is the logical separation of the router’s network operating system (NOS) software from the underlying hardware.  This separation enables the hardware to take on a level of independence from the software.  With an appropriate hardware adaptation layer, the NOS can run on a wide variety of white box hardware implementations utilising merchant packet processing like Broadcom’s StrataDNX chipsets. 

Compare this to the tightly integrated routing approach.  If a white box hardware vendor develops a new hardware platform, cable/MSOs can rapidly take advantage of it. The cable/MSO is not held captive to a limited set of hardware options from a single source.  Network operators also have the flexibility to specify their own hardware or utilise a design directly from organisations like the Telecom Infra Project (TIP) or Open Compute Project (OCP).  For example, Infinera’s 300 Gb/s DRX-30 hardware is available from Infinera, Edgecore Networks, or another third-party supplier that can obtain the design specifications directly from TIP/OCP.

Networking velocity – new silicon faster

Anything that slows the introduction of next-generation packet processing silicon reduces networking velocity.  With tightly integrated chassis-based routing platforms, every generation of packet processing potentially involves a multi-step upgrade, including switching cards, power modules, and fan trays.  In addition, there is always a risk that backplane traces are at maximum capacity, in which case a forklift may be required.  Collectively, these steps create friction – putting sand in the gears of the network and working against the introduction of next-generation packet processing.  Any delay means that cable/MSOs cannot take advantage of the latest packet switching technology – leading to lower density, higher power, and increased networking costs.

Pay-as-you-grow economics

Network operators, including cable/MSOs, report that it has become increasingly difficult to predict traffic demands across diverse network locations. The cable/MSO fiber deep transformation, including virtualisation and centralisation of some functions and distribution of others, only exacerbates this challenge. Uncertainty makes it difficult to optimise working capital and run an efficient network.  Buy too much router capacity and strand precious capex or buy too little and limit or prevent service delivery.   

By delivering self-contained modular “pizza box” routing instances with a leading NOS supplier, cable/MSOs can avoid such issues.  But how can a pizza box scale?  The answer lies in modularity and stackability.  Leading NOS software can interconnect and stack two or three pizza boxes together while enabling them to behave as a single routing instance.  In addition, centralised fabric units can be introduced for additional scalability, enabling virtual backplane interconnects among fabric unit and line unit modules to create a horizontally scalable router ranging from multi-terabit to multiple 100s of terabits in routing capacity.

The modular, “Lego” approach to routing enables pay-as-you-grow deployment and avoids stranding equipment or capital.  It simplifies and accelerates the insertion of new packet processing silicon via the introduction of new pizza boxes. Like servers in a data center, simply replace the legacy module or augment it with a new one.  Because the hardware units are self-contained, power, cooling, and backplane upgrades associated with chassis-based solutions are eliminated.


Overall, cable/MSO HFC networks are evolving to deliver 10G broadband services and introduce new services like 5G. While tightly coupled routers have advanced networks to date, their limitations include restricted hardware choice, reduced networking velocity, and inefficient economics.  By embracing disaggregated routing with white box hardware and industry-leading NOS software that supports single-unit, stacked, and horizontally scalable routing configurations, cable/MSOs can increase their hardware options, improve their flexibility with pay-as-you-grow modularity and economics, and eliminate the sand that is preventing their networking velocity from achieving warp speed.