Improving Azure Customers' Critical Workload Performance with Intel® Optane™ Technology

What’s the Challenge?

Under continuous pressure to deliver top-line growth and meet customer demands for increasingly sophisticated and seamless experiences, an enterprise’s ability to respond quickly to market shifts is fast becoming a core differentiator.

To drive greater business agility, enterprises are now looking to their cloud service providers (CSPs) to provide high-performance infrastructure that accelerates business-critical workloads, while driving down costs.

For Microsoft Azure engineers and solution architects, the challenge now is to deliver solutions that meet these customer performance expectations, for the lowest possible total cost of ownership (TCO).

What’s the Solution?

Integrating Intel® Optane™ technology into the Microsoft Azure cloud enables Azure engineers and solution architects to improve the performance of critical workloads, drive enhanced data center efficiency, and balance costs for enterprise cloud customers. Ultimately, this helps Azure to remain competitive in the marketplace. This article examines some of the specific instances in which Intel Optane technology positively impacts the Azure offering, delivering additional benefits to customers.

Faster Reload Times for SAP HANA Customers

By expanding memory capacity and enabling low-latency access to persistent data, Intel® Optane™ persistent memory (PMem) allows in-memory databases like SAP HANA to load faster, since there is no longer a need to load data from disks or slower storage tiers in the event of system reboot. The faster reload and recovery times may help some deployments to run without high availability for non-production workloads with reduced service windows, and remove clustering complexity and downtimes needed for upgrades and/or patches.

Compared to the reload time on a DRAM-only system, 2nd generation Intel® Xeon® Scalable processors with PMem offer 22x faster reload times for SAP HANA on Azure S224 (Large Instances).1

Platform Consolidation for SAP HANA Solutions

Higher memory densities delivered by Intel Optane PMem enable the scale up or scale out of SAP HANA deployments with fewer Azure S224 SKUs, compared to a larger number of DRAM-only nodes on previous generation processors. This enables Azure SAP HANA customers to consolidate platform footprint, reduce operational complexity, and help lower TCO.

Coupled with PMem, SAP HANA Large Instances offer even higher memory densities at 3-9 TB per socket.1

Increased Data Center Efficiency for Azure Stack HCI

Adding Intel® Optane™ solid-state drives (SSDs) to the cache tier of Azure Stack HCI, plus SATA-based Intel SSDs to the capacity tier, can speed caching and increase virtual machine (VM) density, enabling customers to consolidate workloads onto fewer physical servers, improving resource utilization, and helping to reduce costs.

Intel Optane SSDs increase input/output operations per second (IOPS) by up to 52 percent and deliver up to 26 percent lower latency over solutions using SATA SSDs, while requiring a cluster of three servers rather than four.2

Greater Overall System Memory for Azure Stack HCI

PMem in Memory Mode enables more overall system memory, increasing VM density, dramatically improving resource utilization, and helping lower hardware costs.

By expanding system memory by up to 33 percent with PMem in Memory Mode, Azure can increase the number of VMs by up to 36 percent over a solution with DDR4, lowering the overall cost per VM by up to 24 percent for Azure Stack HCI customers.3 4

For workloads that not only need additional memory, but also need extremely low latency, like high-bandwidth SQL server use cases, Azure can combine PMem in Memory Mode plus App Direct-Dual Mode (called Dual Mode) to boost memory by up to 33 percent, delivering up to 41 percent more VMs per node, and improving IO throughput by up to 76 percent for Azure Stack HCI customers.4

Find Out More

Contact your Intel account team to arrange a demo for your engineering team, and to find out more about how your customers could benefit from Intel Optane technology on Microsoft Azure and Intel.

Or find out more about how Microsoft Azure and Intel are collaborating to deliver a high performance and flexible cloud to meet all customer needs.

Find out more about Intel Optane persistent memory.

Find out more about Intel Optane SSDs.

Notices and Disclaimers

Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.

Performance results are based on testing as of dates shown in configurations and may not reflect all publicly available updates. See backup for configuration details. No product or component can be absolutely secure.

Your costs and results may vary.

Intel technologies may require enabled hardware, software or service activation.

Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy.

© Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others.

Product and Performance Information

1https://azure.microsoft.com/en-gb/blog/next-generation-sap-hana-large-instances-with-intel-optane-drive-lower-tco/.
2Testing by Principled Technologies as of August 7, 2019. For more information, visit, https://principledtechnologies.com/Hpe/Intel-Optane-HPE-ProLiant-Storage-Spaces-Direct-0919.pdf and https://principledtechnologies.com/Hpe/Intel-Optane-HPE-ProLiant-Storage-Spaces-Direct-science-0919.pdf. Added details also on Intel Solution Brief, https://www.intel.com/content/dam/www/public/us/en/documents/brief/nsg-dc-solution-kit-azure-stack-solution-brief.pdf.
3Cost evaluation by Intel as of October 2019. BASELINE CONFIGURATION 384GB DDR4 DRAM: Total storage cost: $5,294; SW cost (per core/ per system): $7,708; CPU cost: $3,788; Memory subsystem cost: $1,416; RBOM cost: $1,300; Total system cost: $19,506; Total cost: 4x $19,506 = $78,024; System cost: 1; Indexed value metric: 1; Indexed value/ $ = 1. PLUS CONFIGURATION 512GB DCPMM (Memory Mode): Total storage cost: $5,294; SW cost (per core/ per system): $7,708; CPU cost: $3,788; Memory subsystem cost: $2,144; RBOM cost:$1,300; Total system cost: $20,234; Total cost: 4x $20,234 = $80,936; System cost: 1.03732185; Indexed value metric: 1.2; Indexed value/ $ = 1.16.
4Testing by Intel as of October 2019. Baseline configuration 384GB DDR4 DRAM: Number of systems: 4; Memory subsystem per socket: DRAM - 192GB (6x32GB; CPU: 2x Intel® Xeon® Gold 6230 processors with 20 cores; Storage description: Number of HDDs/SSDs. PLUS CONFIGURATION 512GB DCPMM (Memory Mode): Number of systems: 4; Memory subsystem per socket: Intel® Optane™ persistent memory (PMem) - 256GB (2x128GB PMem + 6x16GB DRAM, 2-1-1, Memory Mode, 2.7:1); CPU: 2x Intel® Xeon® Gold 6230 processors with 20 cores; Storage description: Number of HDDs/SSDs. BENCHMARK SETUP: Vmfleet test: 18VMs/node; Each VM with 4 cores; 8GB memory; 40GB VHDX; Test file: 10GB. Test setup (735K IOPS: Threads = 4; Buffer size = 4KiB; Pattern: Random; Duration = 300 seconds; Queue depth = 16; 30 percent write; OS: Windows Server 2019 Standard (Desktop) with updated patch. Vmfleet Test: Each VM with 1 Core, 8 GB Memory, 40 GB VHDX. Test setup (1.1M IOPS); Threads = 2; Buffer size = 4KiB; Pattern: Random; Duration = 300 seconds; Queue depth = 16, 30 percent write; OS: Window Server 2019 Standard (Desktop) with updated patch.