HPE self-driving networking push targets AI data center complexity

HPE self-driving networking is being positioned as a core part of Hewlett Packard Enterprise’s AI infrastructure plan, with the company tying data center networking, hybrid cloud automation and AIOps tools into a more automated operating model. Ben Baker, HPE’s director of data center networking, said the effort is meant to reduce operational complexity as AI workloads move more data across compute, storage and distributed infrastructure.

Baker described the current problem as one of scale and visibility: enterprise network teams are dealing with virtual networks, VXLAN Tunnel Endpoints, many protocols and thousands of physical and logical devices. His argument is that operators have too much raw telemetry and too little usable insight, which leaves them reacting to outages and configuration mistakes rather than improving reliability.

The comments were made in an interview with Bob Laliberte, principal analyst for theCUBE Research, on SiliconANGLE Media’s theCUBE. Hewlett Packard Enterprise sponsored the segment, according to SiliconANGLE, which also said sponsors do not control editorial content.

What is HPE's self-driving network strategy?

HPE’s self-driving network strategy combines AI infrastructure networking with AI-assisted network operations. In plain terms, the company wants networks that can spot likely failures, automate fixes and roll back changes before manual work turns into downtime.

The company’s plan has two main parts, according to Baker. One is to make self-driving networks a base layer for what HPE calls the agentic enterprise. The other is to use GreenLake Intelligence, HPE’s agentic AI framework, as a shared intelligence layer across networking, compute, storage, security and cloud infrastructure.

A key integration is Apstra Data Center Director with HPE Morpheus. Apstra Data Center Director is HPE’s data center network management tool, while Morpheus handles hybrid cloud management and automation. Baker said Apstra’s graph database is central to HPE’s approach because it maps relationships among network nodes and helps correlate separate streams of operational data.

HPE is also folding in assets from Juniper Networks and Aruba Networks. Baker said the company is bringing Juniper and Aruba tools together, including expanding Marvis, HPE’s AI-driven networking assistant, into Aruba Central. HPE is also adding support for HPE Networking CX switches to the Mist platform.

Mist is part of Juniper’s AI-native networking portfolio, and HPE is using it in the data center through Mist Networking Data Center Assurance. Baker said the product analyzes network data and connects relevant data types to specific components when diagnosing problems. He also said one of the Mist team’s useful operating choices was linking technical support staff with data scientists.

Why AI infrastructure puts pressure on networks

AI workloads make networking a more visible constraint because models and agents can push large volumes of data between compute, storage and distributed systems. If bandwidth, observability or automation fall short, applications can see higher latency and poorer user experience, according to Baker.

HPE is also investing in hardware for the same reason. Baker pointed to the QFX5250, described as a fully liquid-cooled switch with 64 ports and support for speeds up to 1.6 terabits per second. HPE’s claim is that higher-throughput switching will be needed as AI demand for bandwidth continues to rise.

The company did not disclose customer adoption figures, pricing, deployment timelines or performance benchmarks for the combined self-driving networking effort. For buyers, the near-term signal is portfolio consolidation: HPE is trying to make Juniper, Aruba, Apstra, Morpheus and GreenLake Intelligence operate as one infrastructure management story rather than separate product lines.

This story draws on original reporting from SiliconANGLE.