The demands of AI on network as a service

What is NaaS?

Despite its longevity in the networking industry, NaaS providers can't settle on a unified definition. While NaaS vendors offer similar products, each provider prioritizes different capabilities. As a result, customers don't always have access to the same features.

"NaaS has been misused, bastardized in many different ways," said Suresh Katukam, co-founder and chief product officer at Nile.

The inconsistency is likely due to the varying interpretations of NaaS. Katukam said NaaS isn't like using an MSP or third party to outsource the network. Instead, he likened NaaS to compute as a service. He also compared it to AWS' infrastructure management service that still includes Virtual Private Cloud control.

Despite the different perspectives, however, all NaaS offerings share the same key characteristic: NaaS enables network administrators to manage a network infrastructure without owning physical hardware. Instead of building an in-house network, organizations lease networking services from NaaS providers.

According to Katukam, this enables administrators to focus on microservices -- while still having complete control and visibility -- because of a lack of underlying hardware. Essentially, because NaaS removes the need to manage physical hardware components, network professionals can spend their time optimizing performance. Overall, Katukam said, NaaS providers guarantee consistent network performance.

James Feger, senior vice president of product segments at Lumen, agreed. "In the event of a disaster, having NaaS or network on demand from the infrastructure of a provider … is pretty critical."

Allwyn Sequeira, founder and CEO at Highway 9 Networks, said he sees NaaS as a subset of IaaS. A good NaaS system requires a programmable underlying fabric that can be apportioned to any application based on need, he added.

"What makes NaaS what it is," Sequeira said, "is the context of compute, storage, devices and, ultimately, the new world that AI enables. It'll have to be programmed to meet higher-level objectives."

Meanwhile, Rajarshi Purkayastha, vice president of Tata Communications, said he isn't convinced NaaS truly exists today. For a true NaaS system, it must exist in the network end to end, but that isn't the case. Today, NaaS only performs the specific network functions offered by vendors. Though many of these use cases are helpful, they do not offer performance across the entire network infrastructure.

"It exists in pockets," Purkayastha said. "It needs to be developed end to end from an enterprise architecture perspective."

How AI supports NaaS

The many definitions and changes in capabilities from different vendors make it difficult to pinpoint the areas where NaaS and AI intersect. Added to this complexity are the numerous adoption drivers, use cases and considerations for AI within NaaS.

A significant driving factor to NaaS adoption is fear of missing out, Katukam said. According to Katukam, fear of missing out (FOMO) drives AI trends, including those in NaaS products.

"Every vendor out there is just putting AI on top of every product they built 10, 15, 20 years ago," he said. "[Services and customers] want to get on the AI trend without even knowing what the outcomes they want are."

FOMO isn't unfounded, Sequeira added. The fear is understandable because enterprises that aren't embracing AI risk ending up on the wrong side of history. Because every device is becoming AI-enabled, organizations need to start incorporating it into their networks, he said.

Once organizations decide to implement AI within their networks, they can evaluate the role it could play in a NaaS environment. Important factors to consider include deployment location -- which determines the extent to which AI operates in the network -- and essential use cases, such as data mining.

AI deployment locations in NaaS

According to Purkayastha, it's best to think about AI's effect on the network. Deployment is an easy way to determine AI's effect, specifically in one of the following ways:

• On a device.
• In the office.
• At the network edge.
• In the cloud.

Depending on where AI is deployed, it might not affect the network at all. For example, Purkayastha said deploying AI on a personal device, such as a phone, won't have much effect on a network. In an office setting, it will likely affect only that office, not the whole enterprise. In a cloud environment, however, AI will affect the network from end to end.

Data mining

Data mining is an essential AI use case, as it can help enterprises manage NaaS environments. Many enterprises now deal with petabytes of information daily in their networks, including NaaS networks. These complex networks have become unmanageable, Feger said.

"From a network architecture standpoint, we have to account for all of it," he said. "It goes back to making sure that your architectural decisions are sound."

Security considerations for AI in networking

The panelists discussed security as a major factor in a NaaS environment with AI. However, they discussed AI security from two different perspectives:

1. AI security concerns.
2. AI as a security agent.

AI security concerns

IT professionals have had reservations about using AI, and network professionals aren't the exception. For example, network professionals have increasingly cited shadow AI as a concern, Feger said. However, this is where NaaS can help organizations.

According to Feger, the end-to-end view provided by NaaS enables collaboration among services and partners -- a net good for the industry.

"If we operate in silos, this whole concept of shadow AI is going to become a real problem," he said.

AI as a security agent

While the use of AI in a system can pose security threats, it can also enhance security processes and strengthen an infrastructure. According to Katukam, AI will be able to prevent security threats. Cybersecurity threats increase year over year and happen faster than humans can stop them. AI can prevent threats without human intervention in the future, Katukam said.

Sequeira said he believes AI will be able to further entwine and enhance networking and security practices. He described central IT as the backbone of this practice across different business lines, automated by AI.

"The centralized IT network security backbone -- supporting lines of business with their own top line-driven initiatives -- I think AI is going to play a role in a massive way," Sequeira said.

Drew Geyer, head of IT and cybersecurity at JetZero, said security is a primary AI use case for NaaS. When user identities, network tools and software are known to the network, AI can help maintain a zero-trust policy in an already microsegmented NaaS infrastructure.

AI to minimize human intervention

Eliminating human intervention is one of the promises AI helps NaaS keep, Katukam said. He added that, while AI still has some way to go on this front, it has mostly reached the point where it can reduce human intervention.

Although AI seeks to limit or eliminate human intervention, it's important to consider network engineers when thinking about the future of AI, Geyer said.

"You have to think about the people on your team that are doing the job right now that they won't have to do," he said. "How do you repurpose the people that AI is going to replace? Think about the skills they need, and help them as well."

AI is critical to the future of networking

Most networking experts agree AI isn't a passing fad; it's the future of networking. As networks grow more complex, enterprises must recognize how AI can enable new use cases, improve security and minimize human intervention in manual processes. These use cases, ultimately, enable AI to simplify and enhance network management.

Purkayastha agreed with Sequeira on AI's necessity in networking in the future, specifically regarding the FOMO driving AI adoption in enterprise networks.

"If you're not embracing AI from everything that you're doing, you may as well shut down your business because your competition is going to come and kill you," he said.

He also said it's prudent to consider AI from a broader blueprint perspective. Businesses need to have networks that aren't as rigid as they used to be. From a blueprint standpoint, flexibility helps engineers account for emerging network services.

"The use cases that are going to come in will be influenced not only by what is there today, but what is going to come in," Purkayastha said. He cited quantum as becoming a reality in the next five years, which will change different aspects of the network, including security and AI.

Geyer shared similar sentiments. "You have to imagine what the AI is built on. If you have networking equipment that was designed in 1980, can you really run a full AI stack on that?"

Nicole Viera is assistant site editor for Informa TechTarget's SearchNetworking site. She joined Informa TechTarget as an editor and writer in 2024.