Businesses of all sizes are expanding data centers to support artificial intelligence (AI) and generative AI (GenAI) applications. The next step is upgrading the network that connects it all to locations at the edge — the branch offices, warehouses, stores and other locations where business gets done. Initial AI trials have been generally conducted using the cloud. However, while some enterprises will expand their own data centers to support AI, much of the traffic generated and used by AI application training and processing will come from networks at the edge and be executed using a variety of cloud services.
AI applications generate tremendous amounts of dynamic network traffic, placing enormous pressure on traditional wide area network (WAN) architecture. GenAI applications are changing customer experiences around shopping, coding, personal assistants, video, healthcare, home devices and more. At the edge, GenAI and agentic AI have emerged as business tools that place unique requirements on the WAN. Transporting the vast amounts of data that power these emerging applications among end-user locations, data centers and cloud environments requires an intelligent, sturdy SD-WAN network to operate at the edge and act as a secure on-ramp to multiple cloud data centers.
The Bottleneck Moves to the Network
Cloud computing has required significant upgrades to data center networks, resulting in virtualization and software-defined networking technology. Software-focused solutions that establish dedicated paths across underlying WANs enable businesses to better utilize existing bandwidth to connect distributed locations and access applications in the cloud.
SD-WAN is a software-defined architecture that enables enterprises to connect multiple users with multiple types of devices to multiple cloud environments. SD-WAN ensures accurate routing, threat protection, and capacity management, and simplifies network management. It works across a variety of network infrastructures and is a strong replacement for more expensive MPLS services.
Along Comes AI
GenAI applications are pushing the limits of existing SD-WAN deployments. The low latency demands of GenAI applications, complexity of AI processing and scale of AI architectures require an SD-WAN upgrade. Specifically:
- Extending to the edge: Users of GenAI applications upload high-bandwidth data, such as a video, for analysis. But the response is often lightweight. For example, uploading a 400 MB video and asking for a summary returns a paragraph of text — a few hundred bytes. This upends the traditional model of small uploads and large downloads, straining network architecture. Network delays in processing an AI model can result in incorrect actions, delays and can even stop a GenAI app from working at all.
- Traffic management: AI applications share the network with other business traffic including Software as a Service (SaaS) applications, video calls, email and internet. However, AI traffic often requires access to multiple clouds and underlying networks—especially agentic AI. There is a high volume of encrypted edge traffic that requires low latency and high throughput while other traffic can be delayed or rerouted without error. Rapidly analyzing traffic, applying policies and ensuring security at a level that cannot be effectively accomplished using manual processes requires automation and intelligence applied directly to the network.
- Scale: Beyond the volume of traffic generated by and for AI applications, there is a complexity that must be solved. In addition to variable uploads and downloads, AI applications require rapid access to models and data from numerous sources that constantly change. This data is encrypted, which means that existing methods of traffic profiling are inefficient or outright blind, resulting in delays and errors.
SD-WAN is a valuable network service for businesses of all sizes. As data centers are upgraded to support AI models and operations, the SD-WAN needs an upgrade too. Dynamic traffic management, optimization of encrypted workloads and advanced security upgrades ensure that SD-WAN is ready for AI.
AI-Ready SD-WAN using VeloCloud Robust AI Networking (VeloRAIN)
VeloRAIN is the AI architecture that will underpin the entire VeloCloud portfolio, including VeloCloud SD-WAN. VeloRAIN boosts the power of existing VeloCloud SD-WAN traffic steering and WAN optimization technologies by applying AI and machine learning technology to manage the network. With VeloCloud, a network understands which applications are running on it, AI and not, and can apply customized priority, bandwidth and security at a granular level for each application.
- Traffic management: The ability to identify AI applications, prioritize traffic and dynamically adjust overlay tunnels reduces latency. That ensures the network adequately responds to traffic spikes in both directions, enabling high throughput as needed. VeloRAIN applies AI and machine learning to analyze, prioritize, optimize and secure VeloCloud SD-WAN traffic from the edge of the network to the data center.
- Optimization of encrypted workloads: Not all AI workloads are created equal, but all are encrypted. “Seeing through” the encryption enables the network to understand the correct prioritization of that traffic. VeloRAIN then optimizes use of existing bandwidth without negatively affecting access to SaaS and other cloud services. VeloRAIN aligns traffic steering to individual application flows rather than the WAN underlay to ensure that VeloCloud SD-WAN accurately adapts on demand.
- Security: Security must also scale for AI. VeloCloud SD-WAN has strong built-in security, including an Enhanced Firewall Service that incorporates Intrusion Detection and Prevention Systems (IDS/IPS) to identify and mitigate threats in real time. VeloCloud SASE extends that protection with Symantec SSE for VeloCloud, for threat and data protection when users access any application at the edge, in the cloud or in the data center. This multi-layered approach ensures consistent protection across remote users, edge computing resources, mobile devices, and networks.
Businesses using SD-WAN must leverage the power of AI itself to accommodate the AI and GenAI applications that are and will be coming online. VeloRAIN enables enterprise SD-WAN to identify the presence of AI applications from the edge to the core of the network and optimize the network for the unique requirements of AI traffic. By delivering intelligent insight and automation, VeloCloud effectively manages the WAN for AI and GenAI traffic.