The Enduring Value of IP + Optical


    Jun 29, 2022

    A guest post by Ian Redpath, Research Director of Transport Networks & Components for the global research leader Omdia.

    About the Author

    Ian covers how communications service providers, Internet content providers, and enterprises utilize and deploy leading-edge optical networks, components, and routing technologies.

    Ian talks to Huawei OptiX Live host Musia Yu about the advantages of IP + Optics

    IPoWDM’s promise falls short

    Introducing L0-L3 Network Architectures

    The IPoWDM concept has been around for many years, offering the ability to deploy transmission optics within routing platforms to delayer and simplify networks. Cloud SPs have latched on to IPoWDM for a very specific point-to-point (P2P), short-reach need in their networks.

    Source: Omdia / Figure 1: IP + WDM vs. IPoWDM

    Communications Service Providers (CSP) have picked up on the IPoWDM concept, attracted by its purported network simplification and economic improvements. However, CSPs’ network and operational requirements are far more complex than that of a niche within the cloud SP community. The communications industry has become slightly torn by the allure of IPoWDM, which will succeed in a very narrow, designed-for niche, but remain untenable for many CSP networks and applications.

    CSP challenges and network layer integration

    In contrast to the cloud SP niche (where IPoWDM may work for some cloud SPs), CSPs have more complex network needs.

    CSP optical network requirements include the following:

    • Multiple system reach requirements from DCI, metro, national, and subsea.
    • Many CSP networks are ROADM-rich, meaning wavelengths must pass through many ROADMs.
    • A desire to maximise spectral efficiency.
    • A desire to maximise total system capacity.
    • A need to support a multitude of client side speeds: legacy sub-rates, 1G, 10G, 25G, 50G, 100G, and FlexE.
    • A desire to support maximum line speeds per distance, leveraging programmable modulation formats for traditional and specialised line speeds: 200G, 400G, 600G, and 800G and 150G, 300G, and 500G.

    The devil in the IPoWDM details

    CSPs were attracted to the network delayering and purported TCO savings enabled by IPoWDM. However, once A CSP’s network needs and requirements are factored, the promise of IPoWDM falls apart, leaving CSPs with a multi-variant problem: for IPoWDM to work, very small form-factor pluggables are required. However, to get the pluggables down to QSFP-DD size, key capabilities and performance are sacrificed. To meet CSPs’ true network needs, the form factor size and power draw must increase, but if this happens, the optics can no longer pair with the router host paradigm.

    The IPoWDM concept has been around for almost two decades but has never taken off because of challenges around optimising application performance, power draw, and form factor size. Historically, optics have been too large and consumed too much power, thus devaluing the router’s core capabilities. The challenge continues to this day, as shown in Figure 2.

    Source: Omdia / Figure 2: The “Multi-Variant Challenge”

    CSP networks have distinct needs from cloud data center interconnect (DCI) applications, for which space and power must be preserved and functional integration has a high value. In CSP networks, however, the equation changes in that scale is imperative; optical systems must be massively scalable. IP functionality is optimally delivered with best-of-breed routing capabilities, not muddied by non-core functions.

    While a very high degree of convergence and integration is desirable for DCI applications, convergence in CSP networks would lead to performance penalties.

    IPoWDM demonstrates a similar dynamic in that integration will be valuable in selected network niches, such as its initial design for very short-reach DCI applications. Yet, IPoWDM has turned out to be restrictive in CSP network situations where best-of-breed, functional specialisation, and scale are design imperatives.

    IPoWDM claimed to bring great network economic benefits but has fallen short for the CSP community. The following purported common equipment savings have not materialized:

    • The “cage” for a pluggable will see costs in both a dedicated optical platform and a router.
    • IPoWDM architectures are low cost in only short-reach P2P scenarios.
    • Grey optics are scaled and continue to see rapidly reduced costs.
    • IPoWDM introduces a step function management and control cost.

    IP + optics continues to demonstrate enduring value for the CSP community in the following areas:

    • “The cheapest transponder is no transponder.” CSPs will continue to leverage optical bypass.
    • Major advances in wavelength management continue, both in the core, at scale, and at the edge.
    • CSPs can continue to quickly leverage best-of-breed technology developments.
    • Meets CSPs’ desire to maximise spectral efficiency.
    • Meets CSPs’ desire to maximise total system capacity.

    IPoWDM has begun in a unique market niche, with many design trade-offs at play. For CSPs, the promise of IPoWDM results in disappointment because IPoWDM will not meet the bulk of CSP network requirements. CSPs may deploy IPoWDM in niche situations but scale belongs to IP + WDM and mature and functional ecosystems.

    Further reading


    Disclaimer: Any views and/or opinions expressed in this post by individual authors or contributors are their personal views and/or opinions and do not necessarily reflect the views and/or opinions of Huawei Technologies.


      Leave a Comment

      Posted in


      Posted in