Real-time, high-performance data eliminates challenges of bandwidth, latency, and cloud resources
Non-Volatile Memory Express (NVMe) provides edge computing systems with data center performance, enabling fast data storage and access. The technology is a breakthrough in empowering sophisticated automation capabilities in heavy industrial settings, helping fuel inference computing, artificial intelligence, and machine learning in arenas where environmental rigors have been a roadblock to robust computing. In these settings, fast non-volatile storage enables machine learning algorithms to perform properly even as massive amounts of data are processed.
Consider the intelligent object-detection software used to complete complex visual tasks. More reliable and capable than the human eye, this type of automation is based on deep learning, or neural networks becoming smarter and more efficient as they learn from continued exposure to certain sources of visual data.
While the performance needed to capture and analyze this large scope of data was once made possible only via cloud resources, today’s rugged edge systems blend NVMe storage with CPU and GPU performance, programmed sensors, and software algorithms to unlock advanced automation in more kinds of computing environments. Taking a look at NVMe in action demonstrates how extreme industrial settings can become more competitive by capitalizing on advanced automation.
Shown are the differences between the various interfaces.
Modular NVMe and U.2 add flexibility
NVMe was developed specifically to leverage accelerated storage media like NAND or persistent memory via the PCIe interface. Where SATA SSDs can’t fully capitalize on PCIe lanes, NVMe performs data transfers at double the speed of SATA directly connected to the PCIe lanes. By using the PCIe Gen 3 protocol, data moves along a parallel, low latency path that resembles the architecture of an advanced processor. Newer NVMe SSDs are also now making a transition into the PCIe Gen 4 protocol, providing even faster read/write speeds for SSDs.
CPU cycles are more efficient, avoiding I/O bottlenecks that are common to SATA and fueling the kind of performance required for inference analysis at the edge. This creates a cascade of benefits; for example, read/write capabilities improve with reduced latency, workloads are boosted, and new options for real-time processing are uncovered. Data is stored in flash memory with no moving parts, and the NVMe architecture itself offers the capacity to regulate SSD power. These factors increase reliability, extend battery life, and create positive impact on total cost of ownership.
NVMe also adds flexibility to rugged edge systems via hot swappable capabilities. In an ideal design, an M.2 storage tray serves as a host, allowing quick drive access and replacement. In contrast to a single onboard NVMe drive, hot swappable features ensure flexibility in how data is moved into storage or other master data management facilities.
Optimized systems also support the U.2 interface for connection with NVMe storage. Where small, flat M.2 drives are placed directly on the motherboard, the U.2 cable can access separate 2.5-in. SSDs common to traditional SATA SSDs. Depending on the deployment environment, this may provide a thermal advantage by separating the drive and its generated heat from the motherboard.
Both options communicate to the NVMe storage system’s motherboard directly via the PCIe bus, and both options use the same type of flash memory storage. The U.2 option provides this additional flexibility to end-users, allowing a somewhat higher storage capacity than M.2 (about 8 Tbytes, compared to M.2’s 4-Tbyte maximum). Overall, U.2 enables a faster data pipeline by using up to four PCIe lanes and two SATA lanes.
Benefit to Telematics and Transportation Applications
A massive amount of data is collected by on-vehicle sensors, and that’s only going to increase as 5G proliferates. Supported by AI, this comprehensive, real-time data is ultimately used to improve asset tracking, evaluate driving behaviors, increase productivity, and streamline overall operations and costs. Industrywide, edge devices are also recognized as vital to operations that rely on a mobile workforce.
Because the telematics landscape is quite diverse with a wide spectrum of data, sensor and vehicle types, and operating environments, flexible data operations are key. For instance, strong connectivity is essential for moving vehicles, with telemetry data constantly moving back and forth. When equipped with hot swappable NVMe features, rugged systems in commercial vehicles offer another option for how to send data from vehicles to a central data storage. Drives can be simply swapped out once the vehicle has returned, eliminating downtime and keeping vehicles rolling.
Optimizes the Customer Experience
Rugged edge technology plays a role in everything from inventory management and logistics to personalized shopping experiences and sales forecasting. AI has empowered retailers to construct smart stores—both physically and virtually—as a means of staying competitive in a market where the buyer has all the power. Kiosks with self-service technology must feature rugged designs, blending modular and fanless options for performance in extreme environments and harsh temperatures. Workload consolidation considers compute, storage, I/O diversity, and connectivity to protect performance and enable the cost-saving convenience of effective remote management at the edge.
Public venues and infrastructure facilities can be kept safe with the help of high-performance computing. For example, rugged edge-powered AI lets systems analyze live video automatically or without a human workforce. This contributes to real-time public safety in civic settings such as stadiums, train stations, and airports. Systems deployed here must be able to exist in harsh temperatures (-40°C to 70°C), withstand shock and vibration (5gRMS vibrations and 50G shock), and operate with wide voltage (9 to 50 Vdc).
Coupled with PCIe Gen 3 and NVMe storage, developers are empowered to create applications for data logging, monitoring and surveillance in the most extreme environments. This design approach acknowledges SATA storage technologies as insufficient for this level of edge performance, and instead builds on the same concepts that drive data center design. New rugged edge system designs balance both performance and reliability closer to where data is being generated.
AI/ML at the Heart of Industry 4.0
Industry 4.0 is defined by smart, connected devices in manufacturing, such as high-performing cameras capable of motion detection, signal transmission, and depth perception. Rugged edge systems and their connected nodes process massive streams of data generated by these sensor-enabled devices, powering ever more advanced automation. Data is analyzed in real-time, enabling high-value data interactions that optimize quality control, maximize production, and accelerate time-to-market. Mobile and remote deployments add even more value, capitalizing on the workload consolidation that increases reliability and potentially eliminates on-site tech support.
The rugged edge is where the data is, and today that includes extreme industrial environments that were once limited to cloud computing options. Gartner proposes as many as 25 billion connected edge devices by the end of this year. This represents a new level of hardware intervention, where every type of industrial setting can access the competitive value of more advanced automation.
NVMe plays an important role in real-time data transfer at the rugged edge, value that’s extended through designs that include U.2 options along with multiple NVMe M.2 drives for easy data access, hot swappability, and non-stop performance. These low latency systems avoid data bottlenecks and deliver real-time analytics with the power to improve how an organization operates. Notably, as the NAND flash cost per Gbyte continues to drop, developers have an open path to design an even greater competitive advantage.
Hardware Acceleration at the Rugged Edge
The ability to perform real-time decision making and predictive analytics is an increasingly strategic goal among industrial operations. It’s an engineering challenge that’s here to stay. This new imperative is fueled by rapid digital transformation and a growing appetite for automation upgrades. Rugged edge computing plays a critical and growing role in this landscape, accelerating data processing through sensor input data and allowing access and analytics close to the data source.