Cisco made a big push into the world of intent-based networking this summer, and some are touting it as the next big thing in enterprise networks. In its announcement in June 2017, the firm said this new type of network would “recognise intent, mitigate threats, and learn over time”.
But how does it do this and, more importantly, if an organisation needed it, how would it implement it?
According to Gartner, intent-based networking looks to be the “next big thing” in networking.
In a blog post, Gartner analyst Andrew Lerner says it is not a product or a market, but instead “a piece of networking software that helps to plan, design and implement or operate networks that can improve network availability and agility”.
Amol Phadke, who is global network strategy and consulting practice lead within the Communications, Media & Technology (CMT) industries for Accenture, says that intent-based networking is essentially about the network deterministically following the intent of the operator/operations through the use of automation and analytics.
"Intent-based networking has been around for a while but it is growing in importance given the real-time nature of services and devices that sit on top of the pervasive network,” he says.
According to Kireeti Kompella, chief technology officer of Juniper Networks’ development and innovation team, intent is characterised by the high-level, declarative nature of its statement, and the absence of implementation detail that leads to complexity and inflexibility.
Read more about intent-based networking
He says that at a basic level, you say “what” you want rather than “how” to do it. The implementation details are filled in as the intent is instantiated in the network.
“This is important as it allows networks to be more agile – the implementation can be changed on the fly, in response to changing network conditions,” he says. “Intent-based specification also means that much of the mundane operational tasks can be eliminated, allowing IT staff to work at a higher level, meaning lower costs, fewer errors, greater security, reliability and optimality.”
Ronan Kelly, chief technology officer for EMEA and APAC at Adtran, says intent-based networking is “not a new concept”.
“In fact, it has been in existence and operation in its early forms in operator networks for many years. One of the firms that Adtran acquired almost a decade ago, Luminous Networks, was one of the pioneers in this space, where in a similar architecture to today’s emerging software defined networks (SDN), they had separated control and forwarding plane functions, with the control plane residing on commodity compute resources in the cloud,” he says.
Minimal human interaction required
Kelly adds that in this environment, the human interaction required to build services and circuits across networks with hundreds of network elements, was minimal.
“An intent was entered into the system to build a service from point A to point B, with a particular set of characteristics, and the cloud based computer resources would calculate the available paths that could deliver on the requirements specified, with ultimately, a primary and fall-back path being identified,” he says.
Once the path was chosen, the cloud-based control took responsibility for programming each of the forwarding plane network elements to deliver the service and reprogram them in the event of a network path failure. He says this early implementation demonstrated the potential of an intent based approach.
“The level of detailed interaction with the network elements was greatly reduced if not eliminated, as was the scope for human error.”
The networking industry is full of seemingly good ideas that eventually go nowhere. One has to ask whether intent-based networking is going to help enterprises.
Phadke says that in the enterprise, most companies are looking to simplify their network by utilising SDN-based infrastructures. As their requirements change, they are looking to leverage intent-based networking to manage the infrastructure.
The use cases are diverse – proactive/predictive network-assurance, elastic bandwidth deployment, real-time network topology change, dynamic service orchestration and modifications, he says. “The technology will either help with bottom line benefits, top-line growth, enhance the customer experience, or any mix of the three.”
Phadke sees intent-based networking being used by service providers and, over time, in enterprises as a key tool in the simplification of the network, operational systems and operational processes.
The potential benefit for enterprises will be the ability for them to command network resources from their carrier, as and when they are needed, right down to application flow levels of granularity. “This approach will ensure they are paying for what they need, but not restricted when their needs change,” he says.
Phadke believes the technology will see significant acceleration over the next two-to-three years, “although it’s hard to predict because there are a number of factors as play such as technology maturity, operator pace of transformation, operational readiness and availability of investment”.
Kompella sees the industry “inching toward autonomous networks as we infuse automation frameworks with machine learning and big data analytics”. He says the path to these networks relies on telemetry, automation, machine learning, and programming with declarative intent.
“Take automation for example. Today, we automate topology discovery, path computation, and path installation. We have bandwidth reservation that is responsive to traffic changes, but we need smarter auto-bandwidth, for example, is that traffic spike due to downloading the latest Beyoncé song or a DDoS attack?” says Kompella.
He adds that when networks are self-driving, there will be automatic service placement and service motion; specific upgrades based on configured services; and inductive network response based on machine learning.
But he warned that while significant barriers exist to developing fully autonomous networks, they should motivate the development of intent-based networks. “It took 10 years for self-driving cars to advance from vision to prototype. I don’t expect an autonomous, self-driving network to take this long.”
Kelly says that some of the early implementations of intent-based networking in the datacentre industry are already in place and operational today. A lot of the early initiatives have been spearheaded by open source projects headed up by the likes of Google and Amazon.
“The impacts of this are far reaching where datacentre switching, optical transport, and fixed and wireless access will all be impacted,” says Kelly.
SDN’s role in intent-based networking
Software-defined networking is critical to intent-based networks, says Phadke. It’s the underlying network that needs to be software driven for any of the higher layer intent-based algorithms to be successful.
“As network actions become more algorithmic and deterministic, machine learning plays a key role both in optimising costs and enhancing customer experience. Similarly, advanced automation is important to fully leverage the ‘softwarisation’ of network as well as the operational processes.”
Kelly adds that machine learning coupled with Open SDN control and standardised data models amplify the potential for intent-based networks where insights about both normal and abnormal traffic patterns can be compiled. This permits the optimal operation of the network using the available assets. For example, every network goes through a 24-hour cycle of peak and off-peak traffic.
“When using technologies like NGPON2, the potential exists during off peak time to migrate all optical networking terminal (ONT) endpoints onto a single wavelength, permitting the shutdown of large portions of the access network equipment when it is not needed,” he says. This reduces power drawn, heat generation, and accordingly necessary air-cooling requirements.
Kelly says that with AI implementation, greater intelligence can be applied to such scenarios.
“It will take into account the day of the week, whether it is a holiday period or a major sporting event is on. It considers whether the current weather’s likely to increase data usage, or if there has been a major news event that will drive usage,” he says. “All of this permits real time optimisation across entire networks that wouldn’t be possible without AI [artificial intelligence] and Open SDN control.”
New infrastructure planning
Network infrastructure planning for the 2020 Tokyo Olympics started in 2015, but Kompella says that through an autonomous, self-driving network, planning for events such as the Super Bowl will take only a few days.
“As IT infrastructure is able to order and deliver itself, then self-organises and optimises on site,” he says. “Maintenance will be proactive rather than reactive (as it usually is today), meaning that components and systems will be fixed before an outage happens instead of the failure occurring and causing traffic disruption.”
In the next 12 to 18 months, the intent capabilities embedded into the leading Open SDN controllers will continue to evolve, permitting ever more flexible use cases.
“The most exiting piece is the potential innovation that emerges from outside our industry, thanks to the Open API approach and the use of Open SDN controllers with standardised data models,” he says.
An improvement in network availability and agility could mean intent-based networking being an essential part of an organisation’s digital transformation plans.