AI Implementation Will Reshape Network Architectures and Procedures, Experts Say

The heavy networking and security demands of AI implementation will require revamped architectures and security tools, many experts say — and tools adapted for AI may need to be themselves managed by AI.

The heavy networking and security demands of AI implementation will require revamped architectures and security tools, many experts say — and tools adapted for AI may need to be themselves managed by AI.

"Regarding networking, there are two types of AI — AI for networking and networking for AI," industry analyst Zeus Kerravala wrote earlier this year. "The former uses AI to run the network, and the latter deploys a network to support AI."

Networking and security provider Aryaka touts its Unified SASE as a Service as the optimal architecture for companies that plan extensive use of generative AI, and as Kerravala predicted, AI itself will have a role in overseeing the implementation of the service.

"It's the convergence of networking and security," Aryaka Senior Director of Product Marketing Klaus Schwegler told us. "A from-the-ground-up designed approach to have a unified policy, a unified way of administering, managing, orchestrating policies and have a unified control over such when it comes to networking security."

Optimizing for AI …

To that end, there are two sides of the coin to Aryaka's AI strategy. One side involves tailoring networking and security to the needs of AI, using three optional features for Unified SASE as a Service.

The first, AI>Perform, makes sure that network performance is optimized for AI workloads and applications. The second, AI>Secure, safeguards those AI processes by controlling access and stopping data leakage. Finally, AI>Observe gives users maximum visibility into their AI processes and network usage in general.

"This is crucial in an environment where real-time management and security of networks are paramount, due to the increasing prevalence of AI," wrote Aryaka Chief Product Officer Renuka Nadkarni in a recent company blog post. "The ability to observe and analyze network performance in real time allows enterprises to rapidly identify and respond to potential security threats, ensuring more resilient and robust network operations."

The requirements of AI will reshape network architectures and procedures. For example, because every query to generative AI results in a unique, dynamically generated response, there’s no point in caching data. Yet because traffic to and from servers of AI content will be massive in both directions, extremely low latency is a must.

The solution may be to distribute AI servers geographically, somewhat like SASE points of presence, wrote Orange Group researchers Usman Javaid and Bruno Zerbib a recent piece in TM Forum.

"Future networks must expand cloud-centric architectures toward the edge, bringing LLM closer to data sources, enabling low-latency inference, improving data transfer by processing data locally, while maintaining user data privacy," they wrote.

… and using AI to optimize

The other side of the coin is to use AI to extend security and networking performance in ways that human-controlled processes could not. Schwegler explained how AI could, for example, detect unusual network activity that might escape human notice.

"Using AI tools in order to detect patterns, anomalies," he said. "Traffic behavior that seems abnormal. … All of a sudden, you have a traffic spike, data-transfer rates that you would not detect or too late in order to understand what exactly is happening, who is doing that."

Spotting such anomalies is one of several ways in which AI can boost cybersecurity, according to the U.S. Cybersecurity and Infrastructure Security Agency (CISA). Other potentially AI-assisted aspects of cybersecurity include detecting personally identifiable information (PII) and taking part in forensic examinations.

"There are behavior patterns, trends that can be identified with AI much faster than any other human deterministic machine learning models that can be written," noted Nadkarni in a recent company webcast. "We also need to use AI to enhance the protection capabilities and secure access and secure all assets."

The role of managed service providers

As a company that bridges networking and security, Aryaka sees itself as well positioned to offer its clients and users services that will permit them to maximize their AI usage safely and efficiently. The targeted market is not just direct clients but customers coming through managed-service channels as well.

"MSPs with expertise in AI technologies will play a key role in spotting errors and ensuring the smooth integration of AI into IT workflows," Nadkarni wrote in her blog post. "AI, in turn, will play a crucial role in enhancing real-time network security by providing advanced monitoring, analysis, and error-detection capabilities."

Aryaka Chief Marketing Officer Ken Rutsky said in the webcast that the addition of AI assistance to networking and security tools, and the use of those tools to focus efficiency and delivery toward the best possible AI performance, will usher in a new phase of business opportunity.

"Our goal is to help our customers get it all: performance, agility, simplicity and security without trade-offs," he said. "And as they move into these Gen AI applications, getting it all is going to become harder but even more important and more rewarding."