AlexOakenman - Fotolia
In its most basic form, disaggregated switching is nothing more than network operating system, or NOS, software...
placed on a compatible hardware switch.
It's no different than server virtualization, where Windows or Linux OSes run on standard x86 boxes.
Today, there are a number of NOSes available from vendors both large and small -- suitable for use in a variety of ways, including top of rack, where the Open Compute Project (OCP) has provided the underlying open source switching design standard.
Although disaggregated switching was once touted as a potential sea change for data center networking, most enterprises have stayed away. Instead, large cloud providers and telcos have led the charge, adopting the design to save costs and to support their large-scale infrastructure.
Why are enterprises reluctant to take the plunge? A few reasons stand out, among them:
Knowledge. Disaggregated NOS often requires Linux knowledge, rather than the familiar command-line interfaces known by conventional network engineers. Its deployment may rely on an automation-based Agile process, such as NetOps, which differs from predictable IT processes, like IT service management.
Design. Disaggregated NOSes have an affinity for modern Layer 3 networking designs that differ from traditional Layer 2 enterprise designs. This has led to a mismatch with prevailing technical skills and standardized runbooks of traditional enterprises.
Hardware. OCP network hardware is available primarily from lesser-known white box vendors, and major IT vendors have been generally slow to provide branded OCP hardware, also known as bright boxes.
SDN. There has been a misperception that disaggregated NOS is closely related to software-defined networking. While both approaches can be integrated and often deployed together, one is not necessary to run the other.
The network function is the application. While an x86 server's primary function resides in the application workload, a network device's value is delivered by the NOS itself. Thus, the relationship between enterprise IT and the device is now tied directly to the NOS that enables functions such as the built-in switching or routing residing in the device.
Changes afoot for open source switching, as vendors reassess
Major vendors are now beginning to support disaggregation more broadly, with even Cisco rolling out products that can run other vendors' software. This shift in strategy may reflect demands from suppliers' largest customers, such as telcos and large financial institutions, to emulate cloud providers. As more software and hardware platforms become interoperable, enterprise objections to disaggregation should begin to ebb, as well, and I expect disaggregated NOS will trickle down to smaller organizations. History is on the side of eventual enterprise adoption of open source switching. As the evolution of x86 servers has shown, customers prefer choice.
How can enterprises prepare to gain the benefits of network disaggregation? First, understand why disaggregation is wanted. Which benefits are important? Capex or Opex reduction? Increased agility? A disaggregated system or open source switching platform can offer these advantages, provided one is prepared. The key areas to prepare for and consider include the following:
- Prioritize the benefits, and invest accordingly. Lower Capex is desirable to everyone, but organizations may not have the scale to gain sufficient Capex reduction to justify the increased risk or investments in other areas. I recommend creating a total-cost-of-ownership model before embarking on the change.
- Ensure disaggregation aligns with the rest of the infrastructure. Adoption of modern networking designs, such as leaf-spine or reliance on Layer 3, will affect the rest of the infrastructure. Certify that other teams are ready. Typically, greenfield environments that adopt new technologies will welcome disaggregated networking first. Start slowly before embarking on changing the entire IT network infrastructure.
- Invest in skills. Understand the network automation and training necessary to gain agility to operate modern networks. Staff with NetOps expertise may be hard to find or to train rapidly. Investment costs are hard to quantify easily, so talk to peers to understand their experiences.
- Understand the effects any migration will have on different groups, such as IT purchasing, security or even regulatory compliance. Technical issues alone are not the only barrier to overcome. Gain acceptance from different groups before they get surprised and raise objections.
- Be willing to adopt new technologies. The enterprise needs to embrace new methods and understand investments take time to pay off. If the rest of the organization treats disaggregated networking as an expensive experiment, then a change in attitude is necessary.
- Consider a gradual adoption path. Traditional vendors are adding features such as network automation, merchant silicon and other programmable features in their switches, thus providing some of the same benefits and flexibilities found in disaggregated components. This may be a pragmatic approach to help your organization get familiar with the notion of separating your hardware from proprietary software.