ServiceNow detailed its use of Microsoft’s virtualized, high-performance block storage for AI processing.

The enterprise tech giant leveraged Microsoft Azure Ultra Disk Storage, which is touted by the hyperscaler as the highest-performing tier of Azure managed disks.

Microsoft Ultra Disks are examples of block-level storage volume capable of up to 400,000 input/output operations per second (IOPS) and 10,000 Mb/s of throughput per disk.

ServiceNow employed the tech in its hybrid model setup alongside custom Dell NVMe systems. The firm also leverages Azure’s virtual machines (VMs), namely E-series, M-series, and latest-generation FX-series.

In the case study, growing AI adoption saw the firm needing on-premises performance within the cloud, as matching its NVMe provisions.

According to Charles McDaniels, ServiceNow’s director of Systems Engineering Management for Global Cloud Services, a three-month validation process saw Azure SKUs tested against the company's internal on-premises systems. These benchmarks included load simulations, test performance, and CPU, memory, and latency analytics.

According to ServiceNow, the Azure disks saw results of 99.99% uptime, while supporting the low latency needs of database platforms MariaDB and ServiceNow’s Raptor proprietary PostgreSQL implementation tool.

The business also saw consistency in performance for customers moving to its software-as-a-service (SaaS) solutions on Azure, and has deployed Ultra Disk across 14 Azure regions.

The case study also claimed ServiceNow has reduced usual month-long region implementations to around six weeks.

“We’re excited about the mindset shift this partnership enables," McDaniels explained. "Historically, we purchased and held on to physical gear for years. [We can now] pivot rapidly to newer hardware and virtualization technologies – delivering better performance and agility for our customers.”

The growing adoption of AI has seen storage become an increasing focal point for the hyperscalers. Amazon Web Services (AWS), for one, recently eliminated networking bandwidth limits from storage instances within its Elastic Compute Cloud (EC2).

This week also saw Google Cloud partner with AI storage vendor Vast Data on a hybrid cloud solution to meet rising demand in AI inference and training. The deal will allow GCP users to connect clusters running in the platform with other forms of distributed infrastructure.