Intel® Optane™ DC Persistent Memory Operating Modes Explained

Today, Intel disclosed a new feature of Intel® Optane™ DC Persistent Memory—its two different operating modes. The modes determine which capabilities of the Intel® memory are active and available to software.

To fully understand the modes, let me clarify that servers will be populated with a combination of DRAM and Intel® Optane™ DC Persistent Memory. DRAM has the lowest memory latency. Intel® Optane™ DC Persistent Memory has slightly higher latency, but offers affordable capacity and data persistence.

Memory Mode: Big and Affordable, but Volatile

When configured for Memory Mode, the applications and operating system perceives a pool of volatile memory, no differently than it does today on DRAM-only systems. In this mode, no specific persistent memory programming is required in the applications, and the data will not be saved in the event of a power loss.

In Memory Mode, the DRAM acts as a cache for the most frequently-accessed data, while the Intel® Optane™ DC Persistent Memory provides large memory capacity. Cache management operations are handled by the Intel® Xeon® Scalable processor’s memory controller. When data is requested from memory, the memory controller first checks the DRAM cache, and if the data is present, the response latency is identical to DRAM. If the data is not in the DRAM cache, it is read from the Intel® Optane™ DC Persistent Memory with slightly longer latency. Applications with consistent data retrieval patterns the memory controller can predict will have a higher cache hit-rate, and should see performance close to all-DRAM configurations, while workloads with highly-random data access over a wide address range may see some performance difference versus DRAM alone. Also, data is volatile in Memory Mode; it will not be saved in the event of power loss. Persistence is enabled in the second mode, called App Direct.

Memory Mode seamlessly brings large memory capacity at affordable cost points to legacy applications. Virtualized database deployments and big-data analytics applications are great candidates for Memory Mode.

App Direct Mode: Big, Affordable, and Persistent

Configured in App Direct Mode, the applications and operating system are explicitly aware there are two types of direct load/store memory in the platform, and can direct which type of data read or write is suitable for DRAM or Intel® Optane™ DC Persistent Memory. Operations that require the lowest latency and don’t need permanent data storage can be executed on DRAM, such as database “scratch pads.” Data that needs to be made persistent or structures that are very large can be routed to the Intel® Optane™ DC Persistent Memory. To be clear, if you want to make data persistent in memory, you must use App Direct Mode.

In-memory databases, in-memory analytics frameworks and ultrafast storage applications are good examples of workloads that greatly benefit from using App Direct Mode. Though, through the ingenuity and creativity of our ecosystem, we find cool, new applications for App Direct Mode all the time.

App Direct mode requires an operating system or virtualization environment enabled with a persistent memory-aware file system, including Microsoft Windows Server 2019* and a future update release of VMware ESX v6.7*. We’re also deeply engaged with the Linux community, so please contact your Linux distributor for their release schedule that includes support for Intel® Optane™ DC Persistent Memory.

The modes determine how much memory capacity the OS will register as present in the platform. In App Direct Mode, the DRAM and the Intel® Optane™ DC Persistent Memory are both counted in the total platform memory. In Memory Mode, the DRAM is used as a cache, and does not appears as an independent memory resource, so it is not included in the total memory perceived by the OS. For example, a platform with 1.536 TB of Intel® Optane™ DC Persistent Memory and 192 GB of DRAM would register with the OS as 1.728 TB of total memory in App Direct, but only appear as 1.536 TB in Memory Mode.

Both modes enable Optane persistent memory’s large capacity at a more affordable price per-gigabyte than equivalent DRAM. System administrators can configure the operating modes through the platform’s BIOS or memory management tools, and can even partition the memory pool to operate in different modes simultaneously.

I hope you will start trials of Intel® Optane™ DC Persistent Memory soon through our hardware beta program partners, where it will be available both for server purchases and via cloud Software-as-a-Service offerings. Please join us as we begin the persistent memory revolution!

Click here to learn more about Intel® Optane™ DC Persistent Memory.

Published on Categories Data CenterTags , ,
Alper Ilkbahar

About Alper Ilkbahar

A veteran of the semiconductor and storage industries, Ilkbahar rejoined Intel in 2016 from SanDisk Corporation (acquired by Western Digital Technologies Inc.). Most recently at SanDisk, he was vice president and general manager of storage-class memory solutions, a business unit dedicated to non-volatile, high-performance memory technology. He led the groupd's commercialization efforts, overseeing engineering, marketing and ecosystem development. Earlier in his SanDisk career, Ilkbahar led marketing for enterprise storage solutions and managed several of the companyd's business units, including the Connected and Computing Solutions Group, the Wafer and Components Group and the 3D Memory Group. He joined SanDisk in 2006 with the companyd's acquisition of Matrix Semiconductor Inc., where he had held various engineering and management positions since 2001. Before joining Matrix, Ilkbahar spent nine years in Inteld's microprocessor division, holding design engineering roles and management roles that spanned from the Intel486 processor to the Intel® Itanium® processor. Ilkbahar earned a bachelord's degree in electrical engineering from Boğaziçi University in Istanbul, Turkey; a masterd's degree in electrical engineering from the University of Michigan; and an MBA degree from the Wharton School of the University of Pennsylvania. He holds more than 50 patents in the fields of semiconductor process, device, design and testing and has published multiple conference and journal papers in his areas of expertise.