Mark Berly, the Datacenter Networking CTO at HPE ArubaIn this interview with Mark Berly, the Datacenter Networking CTO at HPE Aruba, we discuss the transformative impact of AI on data centers and the evolving operational model.
As also shared on stage at HPE Discover 2024 in Las Vegas recently, Mr. Berly elaborates on AI’s impact on data centers and how significant it is now and in the future.
“AI is incredibly impactful and will only grow more so. We’re seeing rapid advancements in related technologies, such as interface speeds,” said Mark Berly. “For example, within the next 12 months, we’ll have ASICs capable of 1.6 terabytes per slot. This means compute IO has caught up with compute power, allowing us to process network packets almost in sync with CPU clock cycles.”
Can you give an analogy to help us grasp this speed?
“Sure. Imagine Lake Mead, which holds about 8.5 trillion gallons of water. With a 1.6 terabyte interface, we could theoretically empty Lake Mead in about five seconds. That’s how fast we’re talking.”
What does ‘AI’ truly mean in the context of data centers?
“AI encompasses several facets, including analytics, automation, and augmented intelligence. Many systems today use analytics or automation, but true AI will enable autonomous networks that self-organize, self-provision, self-monitor, self-heal, and self-secure. We’re moving towards networks that can operate independently, drastically improving efficiency and reliability.”
How does this relate to practical use cases in data centers?
“For instance, networks should self-organize once devices are connected, much like how routing protocols have worked for decades. Self-provisioning will handle configurations automatically, while self-monitoring will proactively manage traffic and workloads. Self-healing will predict and mitigate failures before they occur, and self-securing will address the vast majority of traffic that flows east-west within data centers, a significant challenge for current security measures.”
How will these autonomous networks benefit data center operations?
“Autonomous networks will drastically reduce the need for manual intervention, allowing for real-time adjustments and optimizations. This means quicker response times to potential issues, more efficient resource allocation, and enhanced overall performance. Essentially, they will enable data centers to run more smoothly and effectively, adapting to new demands and technologies seamlessly.”
Security is a crucial aspect of data center operations. How can AI enhance this further?
“Last year alone, there were 28,902 critical vulnerabilities identified. AI can interact with databases to identify and mitigate these vulnerabilities in real time, reducing the burden on human operators. This proactive approach is essential as the speed of exploiting these vulnerabilities has decreased dramatically.”
What challenges do we face in achieving these AI-powered data centers?
“The main challenges are gathering and processing telemetry data, managing large data lakes, ensuring security with direct memory access, and most importantly, building trust in these AI systems to make the right decisions autonomously.”
In summary, where do you see the future of data centers headed?
“Networks will continue to get faster and more efficient. AI will play a critical role in their operation, blurring the lines between human and machine interaction. It’s crucial for industry players to start thinking about integrating AI into their data center operations now.”