Microsoft’s Ignite event this week saw the Azure platform integrate more AI into network operations, while Palo Alto Networks seemingly edged out Cisco as agentic security partner of choice for the hyperscaler.

A preview of Azure Boost sees the server subsystem updated to support remote storage throughput of up to 20 Gb/s, and up to one million remote storage input/output operations per second (IOPS).

Designed to offload server virtualization processes traditionally performed by the hypervisor and host OS, Boost now also touts network bandwidth of up to 400 Gb/s for Azure users for both general-purpose and AI virtual machines (VMs).

The Azure Boost update also introduces advanced networking capabilities like remote direct memory access (RDMA) to support low-latency data transit between servers at high speed for AI and HPC workloads.

"New cross-region RDMA ensures reliable, multi-path connections for global AI training, while erasure-coded RDMA adds smart error recovery to keep performance steady even when networks drop packets," explained Group Program Manager of Microsoft Azure Marketplace Max Uritsky.

Microsoft also opened up its virtualization provisions to Nutanix, with the VMware contender supporting Microsoft Azure Virtual Desktop for hybrid environments through Nutanix Cloud Platform.

The integration was touted as expanding the possibilities of virtual desktop infrastructure (VDI) workloads to avoid vendor lock-in, in a possible rebuttal of the ongoing Broadcom and VMware narrative that’s seen the former drive users of the latter into a subscription cloud model.

Both Nutanix and Microsoft were singled out as viable VMware alternatives by Gartner in its new Market Guide for Server Virtualization Platforms, although the research giant stressed that, despite the increase in VM offerings in the market, there still is “no parity replacement for VMware”, with many competitor offerings either incomplete, maturing, or both.

'Palo Agentic Networks'

Another integration announced at Microsoft’s developer-focused summit concerned Palo Alto Networks, which is integrating Azure’s agent factory, Microsoft Foundry, into its Prisma AIRS service.

Adding onto recent AIRS integrations with IBM and ServiceNow, the Microsoft integration sees Prisma AIRS oversee content safety, agentic security, and policy enforcement capabilities across the AI lifecycle.

This specifically includes filtering out shadow AI threats such as harmful content, custom topic detection, prompt injection blocking, as achieved through automated policy checks and real-time runtime protection.

The expansion follows last month’s introduction of Microsoft Agent Framework to the foundry, supplying developers with an open-source software kit for all things agentic AI, as supported by Cisco through enhanced tracing and telemetry for multi-agent systems.

Cisco has AI agent pedigree aplenty, having developed an agentic directory named Agntcy, which it donated to the Linux Foundation earlier this year. On the product front, the networking giant recently updated its Cisco Secure AI Factory with Nvidia to help users accelerate data extraction and retrieval for AI agent models.

Despite that pedigree, it seems Palo Alto Networks has claimed the first platform interconnect against its rivals for direct collaboration on Azure agents.

Outside of collaborations, Ignite also saw Microsoft secure AI security across cloud and on-prem with an expansion of network access suite Microsoft Entra Internet Access.

President of the Identity and Network Access division at Microsoft Joy Chik proclaimed its offering to now be a “Secure Web and AI Gateway,” with a vendor and platform-agnostic service securing, governing, and monitoring network access to any AI model.

The functions echo those found in the Prisma AIRS integration, as well as offering a network file filtering option to analyze file content and metadata in transit, and safeguards against unsanctioned access to model context protocol (MCP) servers by URL.

“With these controls, you can accelerate generative AI adoption while maintaining compliance and reducing risk, so employees can experiment with new AI tools safely,” Chik claimed.