All About AI Ops

Chalan Aras, Senior Vice President at Riverbed discusses how the rise of artificial intelligence is fundamentally reshaping the way we think about the WAN. 

• AI for Networking vs. Networking for AI: How Riverbed balances using AI to manage IT operations (AIOps) with the physical necessity of moving massive amounts of data to "AI furnaces" (GPUs).

• The AI Data Tsunami: Why data is currently growing faster than infrastructure can be upgraded, and why Time is the most critical metric for AI project success.

• The Shift from SD-WAN to "Data Fabrics": Chalan shares a view on why SD-WAN may be fundamentally changing or even declining as the industry moves toward more dynamic, densely connected AI data fabrics centered on the cloud and data center.

• Unified Observability: The importance of moving away from siloed tools toward integrated platforms that can correlate data across desktops, mobile devices, and the core network to remediate issues before users even see them.

The dual challenge: AI for networking vs. networking for AI

Aras distinguishes between two critical concepts: "AI for Networking" and "Networking for AI." Riverbed addresses both, but they serve different functions. "Networking for AI" addresses the physical necessity of moving massive amounts of data to what Aras calls "AI furnaces"—the GPUs that require constant fuel to operate. Conversely, "AI for Networking" (or AIOps) utilizes AI to manage IT complexity. In this role, the system acts as an "IT concierge," identifying the "needle in a haystack" among anomalies and automating remediation so IT teams don't have to hunt for small issues manually.

The AI data tsunami and the currency of time

We are currently witnessing an "AI data tsunami" where data generation is outpacing the ability to upgrade physical infrastructure. Aras notes that "digging trenches and hanging things on poles is hard and expensive," leading to a gap where data grows faster than the network. Consequently, the most critical metric for AI success is time. Because GPUs are "more valuable than gold or platinum these days," having them wait for data is a massive inefficiency. With the rise of Agentic AI, where transactional timings are measured in milliseconds, Aras warns that "Father Time is not with AI," making high-speed data movement essential.

The shift from SD-WAN to AI data fabrics

In a contrarian take, Aras predicts that "SD-WAN is going to die" as a standalone market driver, arguing it is becoming a commodity buried within broader solutions. He suggests the industry focus is shifting back toward the data center and the cloud, where high volumes and complexity are returning. To handle this, Aras proposes a move toward "AI data fabrics." Unlike a traditional network, a fabric is densely connected and "flexes"—modulating itself based on the contours of the traffic. This dynamic architecture is necessary to connect the core to the edge, such as life sciences labs or factories.

The necessity of unified observability

As networks evolve into complex fabrics, siloed management tools become a liability. Aras explains that if a laptop is infected with malware, it triggers alarms across the endpoint agent, the network probe, and the cloud application. Without unified observability, IT teams face three separate alarms. A unified platform correlates this data, identifying that the root cause is on the laptop and allowing for immediate remediation before the user is even aware of the issue.