Broadcom today added an extension to its VeloCloud software-defined wide area network (SD-WAN) platform that makes use of machine learning algorithms and other forms of artificial intelligence (AI) to better optimize traffic flows.
Unveiled at the VMware Explore 2024 Barcelona conference, Broadcom is also adding additional appliances to its VeloCloud portfolio that can scale up to 100 Gbps.
Sanjay Uppal, vice president and general manager for the VeloCloud Division of Broadcom, said the Robust AI Networking Architecture, dubbed VeloRAIN, makes it possible to, for example, optimize network traffic, even when it is encrypted, based on specific application requirements.
New applications will be automatically assigned business priorities using Dynamic Application-Based Slicing, (DABS) to ensure that quality of experience (QoE) per application is maintained across multiple disparate networks, including being able to prioritize network traffic flows for specific end users.
VeloRAIN also provides channel estimation intelligence capabilities for optimizing traffic over wireless links.
Those types of capabilities will be critical as more organizations invest in building and deploying latency-sensitive applications that need to access AI models, noted Uppal. In fact, a report published today by Broadcom finds the top benefit organization expect to gain from investments in edge computing include faster response times for latency-sensitive applications (68%) and improved bandwidth/reduced network congestion (65%).
It’s not clear to what degree organizations will be replacing networking infrastructure but as AI technologies are embedded into infrastructure it should become easier to optimize traffic flows across a wider range of applications. Many of those applications in the age of AI will also be processing more data in near real-time, the results of which will then need to be shared with multiple other applications residing in the cloud and in on-premises IT environments. Each time an application is added to the edge, there is just that much less bandwidth available for all the other distributed applications that need to share the same network.
Exactly how AI will transform the way networking is managed remains unclear, but in addition to eliminating many manual tasks it should become a lot more feasible to build and deploy highly dynamic applications in a way the ensure bandwidth resources are always available. Ultimately, the overall level of networking expertise required to support those applications should be substantially less than what would otherwise be required today.
Broadcom, of course, is not the only provider of networking infrastructure investing in AI and there may even come a day when the AI models and agents that are automating various tasks will need to interoperate with one another. After all, most application services will eventually need to be integrated across multiple organizations that are unlikely to have standardized on the same networking infrastructure.
The one thing that is certain, however, is the level of abstraction at which networking is managed is about to become a lot higher. The challenge, as always, will be to make sure there is still enough human expertise available to resolve the inevitable issue that an AI agent isn’t going to be able to do on its own.
