“Things that came apart could be put together again, but never exactly the same.”
― Deb Caletti, The Six Rules of Maybe
Despite the recent hype around the disaggregated network, data centers have long separated their control plane software from their hardware, as well as mixed and matched vendor products. Disaggregation in transport networks, however, has only recently gained adoption among service providers. To clarify, “disaggregation” means physically separating a traditional optical transport system into functional modules, instead of the traditional large shelf (chassis) populated with line cards to support functionality such as transponders, muxponders, switching, and ROADMs.
Inefficient designs create constraints
This chassis-based approach is inefficient for both operator and equipment manufacturer. From the manufacturer standpoint, when we plan and design a chassis-based system, we immediately constrain innovation of that product for its entire lifetime. The chassis design determines important characteristics such as line card physical size, available power, thermal dissipation, and backplane speed. If a new technology becomes available after the chassis design is in place, and that new technology requires a different slot size, more power, or a faster backplane, the only option is to design yet another chassis. This increases cost and constrains choice and flexibility for manufacturer and operator alike.
Additionally, whenever a new line card is developed for a chassis-based system, it must be regression-tested against the other cards in the system to ensure compatibility. This process increases implementation cost and prolongs time to market for the manufacturer—who ultimately passes this cost on to the operator via price increases.
By disaggregating line card functions into blades, it becomes possible to develop these almost independently of each other, increasing the pace of innovation and cutting costs overall. By eliminating the chassis and focusing on functional components, manufacturers can design modular units without concern for the power, space, or backplane performance requirements of other elements.
In addition to the cost and time burdens of chassis-based equipment, these designs scale poorly. They tend to result in uneven costs, notably high upfront investment in shelf, power supply unit, and other common elements, with revenue-generating line cards deployed over time. Disaggregating functions into 1RU components enables an operator to even out the costs and purchase units to support immediate demands and address future needs incrementally. That’s the defining benefit of the pay-as-you-grow approach.
The technical advantages of disaggregation include lower costs, faster deployment, and efficient scaling. But they don't mean much if your job is to plan, build, or operate a network. In fact, disaggregation presents challenges that can make things harder from an operational perspective. Let’s take a look at a couple of these challenges.
The power dilemma
Disaggregation proliferates individual units that each need their own source of power. Most telecommunication facilities use distributed -48 V DC plant for power, and an electrician is required to furnish power access for new chassis installations.
When systems were chassis-based, this was not a major concern; an electrician could wire the chassis, and telecommunications technicians could add line cards later—without needing additional work by the electrician. With a disaggregated system, however, every new blade added means the telecom technician must coordinate with the electrician again. It’s easy to see how this might cause cumulative delays due to demands on the electrician’s schedule.
There are a couple of options for mitigating this. Some manufacturers offer housings: simple, integrated power access devices with slots for the blades. These can be prewired for blades to be added later. Another option is “connectorized pigtails” that can be plugged into blades. Since these connectors are not exposed like traditional spade connectors, the electrician can prewire power drops and safely leave them disconnected until needed, without fear of shorting. The technician then snaps the connector into the blade as it is installed.
The granularity dilemma
Another concern is the consequence of breaking up fewer large elements into many smaller ones. Instead of managing a consolidated chassis and cards as one unit, the operator must now manage each blade individually. This “granularity problem” makes management difficult and it gets worse as networks grow.
To solve this, manufacturers have developed software that takes advantage of SDN and software container technology to logically combine multiple blades into a single managed system. While these software controllers provide similar functionality to traditional network or element management systems (EMS/NMS), they also offer functionally based insight rather than merely physical network views. This affords greater flexibility, because a functional view can be configured to the specific structure and work practices of the operations teams (rather than having to match operations to the functionality of the EMS/NMS).
Worth the effort
Platforms like these let operators take advantage of disaggregated architectures, while minimizing adverse effects on operations. If implemented appropriately, they can even reduce effort and improve efficiency. Evolving to a disaggregated architecture may seem complex and perplexing at first, but with proper planning the rewards are worth the effort and the challenges can be solved collaboratively.
Bill Beesleyis responsible for MSO Strategy and Planning for Fujitsu Network Communications, Inc. He has more than 20 years of telecommunications industry experience in a variety of executive engineering and leadership roles. Bill holds a Bachelor of Science in Information Technology Management, and a Master of Applied Science in Telecommunications Management.