Over 90 percent of enterprises are in the midst of their digital transformation journey2 to become data-centric businesses. Thus, they will need to capture, analyze, and secure increasing amounts of data.
Data growth is accelerating the demand for performant, data-intensive computing. As the demand for compute grows, memory capacity in the system typically needs to scale along with it.
Large pools of DRAM help accelerate computing with low latency, but DRAM is limited in capacity, is volatile, and expensive. Indeed, DRAM is becoming one of the highest priced components in modern server bills of materials. Alternatively, block storage is large, cheap, and persistent, but slow to bring data to the CPU.
Intel® Optane™ persistent memory (PMem) bridges the gap with an innovative memory technology that delivers a unique combination of affordable large capacity and support for data persistence. With 3rd Gen Intel® Xeon® Scalable processors and Intel Optane PMem 200 series, organizations can optimize capacity, performance, and cost of their workloads by enabling tiered memory and improved tiered storage.
PMem is compatible with 2-socket and 4-socket platforms and is helping to turn more data into actionable insights.
Optimize Workload Performance and Reliability with Intel Optane Persistent Memory 200 Series
Both DRAM and Intel Optane PMem 200 series sit on the DDR bus and can be used together, complementing system memory and enabling high-performance, tiered memory architectures. Available in 128 GB, 256 GB, and 512 GB modules, it offers both large capacity and persistence. PMem 200 series delivers an average of up to 32 percent more memory bandwidth than the previous generation. These capabilities:
- Enable tiered memory and storage architectures that help improve overall system performance, cost, and efficiency
- Help accelerate large-memory computing by keeping more data closer to the CPU
- Accelerate restart times with reduced I/O by persisting data in memory and not requiring data reloads from storage
- Reduce power consumption for large-memory nodes
PMem brings important advantages to data-intensive, compute-intensive, and capacity-demanding workloads. From cloud to databases, in-memory analytics, virtualized infrastructure, content delivery networks, and more, these architectures can easily take advantage of large-scale, tiered memory using Intel Optane PMem. Enabling large memory pools with Intel Optane PMem helps accelerate time to insights, cost savings, and new revenue.
A Powerful Foundation for Optimized Platforms
3rd Gen Intel Xeon Scalable processors and Intel Optane PMem 200 series create a powerful foundation for optimizing platforms for multiple types of workloads. With 3rd Gen Intel Xeon Scalable processors for 2-socket and 4-socket platforms, each memory channel can support one Intel Optane PMem 200 series module (six channels on 4-socket platforms and eight channels on 2-socket platforms). Each module draws only a maximum 15 watts of power, offering high capacity without high power demand and creating large-memory platforms to power data-centric businesses.
Tiered Memory and Storage
Intel Optane PMem offers system architects and application developers new options to create hierarchical memory and storage tiers to address workload challenges. A tiered approach can better optimize the resources of the platform for data access and transport. Programmers can leverage the speed and proximity of technologies closest to the CPU, while taking advantage of the overall capacity available in the system.
Tiered memory – For a tiered memory model, low-latency DRAM offers fast access to local memory, while Intel Optane PMem 200 series creates large capacity to store and protect data with DRAM-like speed. Depending on the application, PMem can be persistent or volatile.
The combination of DRAM with persistent memory:
- Allows fast in-memory computing on massive data sets
- Allows for consolidating more virtual machines on a platform in a virtual environment with large, performant capacity
- Reduces risk of losing critical data when the application is persistent-memory aware
- Speeds time-to-solution from large computations where intermediate results can be persisted and reloaded for final analysis
Tiered storage – With a tiered storage model, Intel Optane PMem 200 series creates a performance tier, delivering fast, byte-addressable access to most frequently needed data. Other technologies, such as SSDs, offer slower access to warm data storage in a capacity tier where speed of access is less critical.
Tiered memory and storage architectures help optimize speed, latency, capacity, and cost. Choosing the right combinations for each application can help optimize systems and their workloads.
Affordable Large Capacity
When deployed as tiered memory to augment DRAM, Intel Optane PMem 200 series enables consolidation and reduced server footprint. These lead to lower software licensing costs, reduced power consumption, and other operational efficiencies.
In-memory and large data computing – Intel Optane PMem 200 series supports large in-memory databases with fast access to more data, without loading it from storage. Workloads processing massive data sets, such as scientific or data warehousing and analytics, can work continuously without repeatedly loading and storing data locally.
Virtualized infrastructure – Intel Optane PMem can offer greater memory capacity per socket than DRAM for virtualized infrastructures. PMem leaves more headroom for virtualizing future workloads that might require large memory capacity, eliminating the need to run them on bare metal.
Drive Application Innovation and Explore New Data-Intensive Use Cases
With Intel Optane PMem 200 series, developers have new opportunities to optimize tiered memory and storage for their applications. They can drive innovation and capabilities using the same persistent memory programming model introduced with the first generation of PMem.
Rapid adoption is easy, and customers are able to take full advantage of PMem capabilities with a growing global ecosystem that includes the following:
- ISVs and OSVs
- Virtualization solution providers
- Database and enterprise solution vendors
- Data analytics vendors
- Open source solution providers
- Cloud Service Providers
- Hardware OEMs
- Standards bodies, such as the Storage Network Industry Association (SNIA), ACPI, UEFI, and DMTF
Seamless Migration from Previous Generation
PMem 200 series is compatible with the software ecosystem already established for previous-generation Intel Optane PMem 100 series. Migrating to or adding new systems built on 3rd Gen Intel Xeon Scalable processors with Intel Optane PMem 200 series is transparent to software designed for the previous generation.
Aerospike 2.5X3
Katana Graph 2X4
VMware 25%5
CDN 63%6
Intel Optane PMem 200 Series Addresses Key Data Center Challenges
Intel Optane PMem 200 series addresses many of the computing challenges that data centers face today. These challenges include high DRAM costs for large-memory nodes, data protection during outages and maintenance, emerging workloads that take advantage of tiered memory and storage architectures, and more.
Active Workloads with PMem Deliver Real Value Today
Many system solutions already take advantage of PMem, leveraging capacity and persistence to enhance workload performance and reliability across multiple use cases and domains.
Programming Model
The software interface for Intel Optane persistent memory was designed in collaboration with dozens of companies to create a unified programming model for the technology. This software interface is independent of any specific persistent memory technology and can be used with Intel Optane PMem 200 series or any other persistent memory technology. The Storage Network Industry Association (SNIA, https://www.snia.org) published a specification of the model.
The model exposes three main capabilities: management, storage, and memory map paths.
- Management path allows system administrators to configure persistent memory products and check their health.
- Storage path supports the traditional storage APIs, where existing applications and file systems need no change. They simply see the persistent memory as very fast storage.
- Memory-mapped path exposes persistent memory through a persistent memory-aware file system. Applications have direct load/store access to the persistent memory. This direct access does not use the page cache like traditional file systems. It has been named DAX by the operating system vendors.
The Persistent Memory Development Kit (PMDK), available at pmem.io, provides libraries meant to make PMem programming easier. Software developers only pull in the features they need, keeping their programs lean and fast on PMem. These libraries are fully validated and performance-tuned by Intel. They are open source and product-neutral, working well on a variety of PMem products. The PMDK contains a collection of open source libraries, which build on the SNIA programming model. The PMDK is fully documented and includes code samples, tutorials, and blogs. Language support for the libraries exists in C and C++, with support for Java, Python, and other languages in progress.
With the industry-standard persistent memory programming model and downloadable PMDK, developers can build simpler and more powerful applications today to benefit operations now and into the future.
Operational Modes
Intel Optane PMem 200 series has multiple operating modes:7
Memory Mode delivers large memory capacity without application changes or system modification and with performance close to that of DRAM, depending on the workload. In Memory Mode, the CPU memory controller sees all of the Intel Optane PMem 200 series as volatile system memory (without persistence). The CPU uses DRAM as a fast cache to Intel Optane PMem. In Memory Mode, data in the modules is protected with a single encryption key that is discarded upon power down, making the data inaccessible.
Memory Mode’s large capacity allows workloads that are constrained by memory capacity to operate more efficiently. Virtual solutions can also scale easier on a platform, with more utilization from the same server, helping to reduce TCO.
App Direct Mode enables large memory capacity and data persistence. Software can access DRAM and persistent memory as two separate pools of memory. In App Direct Mode, software and applications that are enabled for the industry-standard persistent memory programming model can talk directly to PMem. Direct access reduces the complexity in the stack and takes full advantage of byte-addressable persistent memory with cache coherence, ensuring multiple copies of the same data across the node are in sync. (See https://www.minitool.com/lib/cache-coherence.html.) This capability extends the usage of persistent memory outside the local node and provides consistent low latency, supporting larger datasets.
App Direct Mode can also be used with standard file APIs to access persistent memory address space (called Storage over App Direct). Access is done without modifications to the existing applications or the file systems that expect block storage devices. Storage over App Direct presents Intel Optane PMem as high-performance block storage, without the latency of moving data to and from the I/O bus.
In App Direct Mode, data is encrypted using a key stored on the module in a security metadata region, which can only be accessed by the Intel Optane PMem 200 series controller.
The modules are locked at power loss. A passphrase is needed to unlock and access the data. If a module is repurposed or discarded, a secure cryptographic erase and DIMM over-write operation keeps data from being accessed.
Data at Rest Secured for Enhanced Protection
With Intel Optane PMem 200 series, data at rest is better protected by strong, industry-standard security measures. All data is encrypted by 256-bit Advanced Encryption Standard (AES-256). With hardware-based cipher key processing, encryption is transparent to application software. Thus, no software code changes are needed when adding PMem. Hardware-enabled encryption on 3rd Gen Intel Xeon Scalable processors results in strong, industry-standard data security with low impact on performance.
Data Persistence and Automatic Cache Flushes
Unlike DRAM, Intel Optane PMem 200 series, when operating in App Direct mode, retains data during a planned or unplanned restart, avoiding time-consuming data reloads. Maintaining data in PMem, means less down time, fewer losses from system outages, and increased operational efficiency.
Applications that manage data structures in persistent memory routinely call cache-flushing commands to move data stored in the CPU cache to persistent memory. Waiting for these cache flushes to complete can reduce performance. Applications can now avoid those waits altogether with extended asynchronous DRAM refresh (eADR). eADR is a new platform feature available with Intel Optane PMem 200 series with 3rd Gen Intel Xeon Scalable processors. The application automatically detects eADR and skips cache flushing, even if it experiences a system crash or power failure. This means application “Stores” are considered persistent as soon as they are visible to the application. This ability brings “lock-free” programming to persistent memory. eADR helps free up resources and improves performance.
Using eADR with Intel Optane PMem requires platform hardware support. The system requires additional stored energy available to allow cache flushes to occur on power failure. Coupled with application checking for the enablement of eADR and following the persistent memory programming model will ensure a solution is enabled fully for eADR support.
Data Utilization
Intel Optane PMem 200 series is the next-generation of a groundbreaking memory technology innovation. Deployed with 3rd Gen Intel Xeon Scalable processors, this technology can transform critical data workloads—from cloud to databases, in-memory analytics, virtualized infrastructure, content delivery networks, and more.