Data-on-Demand: The Gap that Only Intel Persistent Memory Can Fill
Mar 26, 2021
Today's users are demanding more and more from their applications. The trend to move applications of all kinds online has been in place for years, but as you can guess, the worldwide pandemic served to bend the curve further upward. As users demand more of their applications, software providers are forced to demand more of their servers and storage. For many companies, Intel® Optane™ DC Persistent Memory is the key to answering this demand.
One of the most innovative advances in memory technology to hit the market in years, Intel® Optane™ persistent memory delivers a combination of affordable large capacity and support for data persistence. It is ideal for a wide range of use cases and can provide solution providers quicker insights from data-intensive applications.
Among its benefits include native persistence and the ability to store data even when powered off, two operational modes allow customers the flexibility to take advantage of Intel® Optane™ Persistent Memory benefits across multiple workloads, large capacities at lower than DRAM prices, and 256-AES hardware encryption to ensure the security of your data.
Three Factors Driving Data Demand
In practical terms, global application demand has created three interlocking trends. First, more core computing power and multi-thread capability are needed. In-demand applications need faster, more powerful processing to grow.
Second, as a result, servers are requiring more and more memory. By 2025, the average server will need approximately .5 Terabytes worth of information. The more users do with applications, the more data is required.
Thirdly, despite its speed, DRAM will have a hard time keeping up with this level of demand. Compared to other technologies like persistent memory, it is too expensive and too volatile.
Older Storage Modes are Cheaper, But Not Practical
On the other end of the spectrum, older technologies such as SSD, disk, and tape, although relatively cheap, won't be able to deliver the performance necessary to meet the bandwidth and latency needs of today's servers ( much less tomorrow's).
Fortunately, Intel Optane DC Persistent Memory (PMem) has stepped in to fill the gap between RAM and storage. Let's face it, as bandwidth requirements increase, your applications will need to process storage-size chunks of data with RAM-sized latency.
Bandwidth Vs. Latency
At this point, it's crucial to understand the difference between bandwidth and latency and why both are important. Bandwidth is the volume of data that can be sent or received at a given time. Think of it as the number of lanes on a highway. Therefore, the higher the bandwidth, the better. Latency refers to the amount of time it takes for the data to travel through a given system. Think of it like the route the data takes - the shorter the course, the sooner the data arrives. Therefore, the lower latency, the better. Later we'll look at what types of applications require high bandwidth versus which require low latency.
Persistent Memory - The Best of Both Worlds
Intel Optane Persistent memory allows for data access as quick as RAM but stored in sizes comparable to disk. While it operates at a slower speed than DRAM, it makes more data available to an ever-faster generation of CPUs.
Not that traditional RAM isn't important. Fortunately, Intel's Xeon processors have received a BIOS that allows the user to control how RAM and Persistent Memory can be used in three distinct modes:
- Bandwidth-optimized - provides for the largest throughput of data to the CPU
- Latency-optimized - minimizes the latency of the DRAM data
- Balanced - allows for DRAM to be used as cache and PMem to be used as main memory while still providing DRAM-like performance.
Persistent Memory - Business Benefits
Sound interesting? Optane Persistent Memory isn't just a concept. It's hard at work in many large corporations' servers providing benefits in the areas of analytics, databases, and data center infrastructure.
As your organization grows, the more critical, timely, and accurate business analytics become. Unfortunately, the more complex business data queries become, the more information they require to process. By providing rapid CPU access to more significant amounts of data than ever, PMem allows companies with applications like SAP HANA to run fewer software instances on fewer nodes. These efficiencies not only reduce the cost of ownership but also increase overall performance.
Beyond business analytics, PMem enables companies to accomplish more in real-time, thanks to instant database access. To support online bidding, fraud detection, and digital payment, financial institutions require the quickest, most reliable data possible. PMem supports these needs by opening a larger pool of data for instant CPU access and providing large amounts of non-volatile storage to support immediate disaster recovery. For example, Aerospike, a popular provider of real-time computing data, leverages PMem to extract deeper and more-reliable insights for its customers.
Infrastructure and Virtualization Benefits
For cloud providers, optimizing their investment in infrastructure can be crucial. Limited data access per virtual machine can result in costly hardware investments. After all, with traditional DRAM, there's only so much data available to a CPU at any given time. PMem enables the hottest, most important data instant CPU access on a large scale. Therefore, with each core being able to process more data, each server has the potential to run more virtual machines. For example, Asiainfo, a telco software provider, achieved latency reductions of up to 68% by giving their customers access to more data stored in memory versus SSDs.
Need to Make Your Data Persistent?
UNICOM Engineering works diligently to prepare for the transition to next-generation platforms and the launch of enabling technology as both an Intel Technology Provider and Dell Technologies Titanium OEM Partner. Our deep technical expertise helps drive shifts to the latest server platforms. We are flexible and agile in our pursuit to bring your solutions to market. Our global footprint allows your solutions to be built and supported around the world by a single company. Now you can begin to understand why leading technology providers trust UNICOM Engineering as their trusted deployment partner. Contact us for more information by visiting our website and scheduling a consultation.