Microsoft touts bigger, faster Azure VMs as data deluge grows

NVMe storage should mean better performance for data-intensive workloads, says tech giant

Microsoft is rolling out two larger memory-optimized Azure instances that offer faster response times and throughput for remote storage operations, making use of both Azure managed disks and Intel's Xeon "Ice Lake" processors.

The software giant previewed the NVMe-enabled Ebsv5 VM sizes at the company's Ignite 2022 show in October. Microsoft is aiming the instances at data-intensive workloads at a time when data generation continues to accelerate, forcing enterprises to rely on more compute power and storage capabilities to process and analyze it all.

Those workloads include memory-intensive business-critical applications, relational database servers, and in-memory data analytics.

Redmond this week announced the general availability of the NVMe-enabled Ebsv5 and Edbsv5 VM series and added two new sizes, the E96 and E112i, to the family. The new Ebsv5 and Edbsv5 NVMe VMs offer up to 260,000 IOPS (input/output operations per second) and 8,000MBps of remote disk throughput.

They also include up to 672GiB of memory and – for the Edbsv5 VM series – up to 3,8000GiB in local SSD storage. The Ebsv5 VM series doesn't include local storage or support standard SSDs and HDD disk storage.

They're powered by Intel's Xeon Platinum 8370C chips and include such Intel technologies as Turbo Boost 2.0, Hyper-Threading, and AVX-512.

What all this means for enterprises looking to squeeze more business-critical information from the massive amounts of data they're generating is they can consolidate existing workloads into fewer or smaller VM sizes. The end result is savings on infrastructure and commercial software licenses, according to Priya Shan, senior product manager at Microsoft.

The new NVMe-enabled sizes are available in 13 regions: North Central US, Southeast Asia, West Europe, Australia East, North Europe, West US 3, UK South, Sweden Central, East US, Central US, West US 2, East US 2, and South Central US.

Microsoft will add more regions in the future, Shan wrote.

Along with the new larger sizes, Redmond also is bringing NVMe to existing smaller sizes – E48 vCPU and E64 vCPU – which, we're told, will drive higher performance. They're being offered at no additional cost.

The high-performance NVMe storage interface is faster and more efficient that other protocols, including SCSI, which until now was the only option in most Azure VMs. Microsoft in April 2022 launched the Ebsv5 and Edbsv5 VM instances, delivering up to 120,000 IOPS and 4,000 MBps of remote disk storage throughput.

All but the E112i will offer the option of either NVMe and SCSI. The E112i will only come with NVMe.

Adding NVMe support allows organizations to leverage the Azure VMs to increase both higher VM-to-disk throughput (up to three times higher) and IOPS performance per core (up to six times higher) compared to the Ev4 generation.

"Customers usually deploy data-intensive workloads such as online transaction processing systems, data warehousing, and analytical applications on memory-optimized Ev5 VM series," Shan wrote in October last year. "While the Ev5 VMs meet the performance requirements for many business-critical applications, some large on-premises database environments require even higher VM-to-disk throughput and IOPS performance per core, which the latest NVMe Ebsv5 VM sizes can now support."

This comes as Microsoft continues to chip away at AWS's dominant share of a global cloud infrastructure services market that in the first quarter saw enterprise spending blow past $63 billion, over $10 billion more than in Q1 2022, according to Synergy Research Group.

During those three months, both Microsoft and Google saw their market share grow a percentage point to 23 percent and 10 percent respectively. AWS's share remained steady in the 32-34 percent range. In all, the top three accounted for at least 65 percent of the overall market. ®

More about

TIP US OFF

Send us news


Other stories you might like