Top 10 memory & storage considerations for IIoT applications
Author : Arthur Sainio | Director of Product Marketing | SMART Modular Technologies
01 August 2020
SMART_nuclear plant_580x280
As Industrial Internet of Things (IIoT) hardware systems continue to evolve, the amount & quality of data generated is increasing exponentially. Accordingly, the movement, storage & analysis of IIoT data is also changing.
This article was originally featured in the August 2020 issue of EPDT magazine [read the digital issue]. Sign up to receive your own copy each month.
There is no prescribed data hierarchy, as there are many potential data sources and many data clients. The vast amounts of data are not limited to localised connections serving purpose-built applications, but are available across a network of connections. Here, Arthur Sainio, Director of Product Marketing at legacy, custom & specialty memory & storage solutions provider for electronics, computing, communications & industrial OEMs, SMART Modular Technologies explains why, within the IIoT ecosystem, electronics design engineers should consider what memory and storage technologies should be used in IIoT hardware.
Connected factories and automated machinery, smart buildings, the smart grid and smart agricultural equipment, connected transport and healthcare applications are all part of the evolving fabric of the Industrial Internet of Things (IIoT). And as IIoT applications continue to gain broader adoption within core industries including manufacturing, construction, energy, transport, agriculture and healthcare, memory will become an increasingly important part of the core fabric of the IIoT. However, the memory used in IIoT requires unique characteristics to ensure it is reliable, even under often challenging operating conditions. Dynamic random access memory (DRAM) and flash memory are both used in IIoT applications, based on factors including unique data characteristics, endurance and reliability for 24/7 operation, performance, power consumption, longevity, scalability, ruggedness, temperature, security and cost.
In this article, we will review the top 10 memory and storage factors electronics design engineers should consider when developing IIoT applications:
1. Data characteristics
Figure 1. An example of a DDR4 memory board used in IIoT embedded systems
IIoT data characteristics are a key metric in determining the type of storage media, as well as the configuration of the memory, that is used in a system’s design. IIoT systems that collect video data, for instance, can leverage mainstream 3D NAND flash technologies, such as triple-level cell (TLC) or quad-level cell (QLC), which are used in various types of SSDs and SD cards. When capturing video data, not every bit of data is critical, so the flash media can be consumer-grade. This would apply to DRAM technology as well: video data does not require enterprise-grade DRAM or error correction code (ECC), therefore IIoT systems capturing video data can leverage consumer-grade DRAM, which is either down-board or in a module form factor. However, with numerical IIoT data, it is more important to maintain data integrity. Down-stream data analysis or monitoring may require alerts if something is wrong: for example, a system relying on a temperature monitoring alert if one bit of data is outside required parameters. In these situations, every bit of data counts; therefore, the type of flash media should be considered more closely to ensure critical data is not lost.
In regard to DRAM, it would be worth considering the addition of the ECC function to the system design to minimise any DRAM bit failure issues. ECC functionality is available in single DRAMs (ECC DRAM) or in a module form factor.
2. Endurance & reliability for 24/7 operation
Typical IIoT systems are designed for long-term continuous operation. Twenty-four-hour operation, seven days a week is very common for outdoor signage, factory process monitoring and building automation applications. Table 1, comparing the types of available flash media and various P/E cycles, should be referenced when selecting the appropriate flash media type. The system design, with respect to flash, needs to match the use-case: how often data will be written to the flash and how long it needs to be stored. The more critical the data, the more care needs to be taken in selecting the flash type. Many industrial systems, for example, are still using industrial-grade SLC CF Cards. These highly reliable (and expensive) flash cards were originally designed for cameras and are rated for 100,000 P/E cycles. In many cases though, standard SD cards or MicroSD cards can fulfill IIoT data storage requirements.
DRAM has unlimited endurance, so there are no issues with the number of times data can be written to or read from the DRAM. The core consideration when selecting DRAM is reliability. For high reliability IIoT systems, it is desirable to use enterprise-grade DRAM that has undergone more robust testing to reduce the chance of future bit failures in the field. This is in contrast to using low cost consumer-grade DRAM. If an IIoT system used for a critical application is not designed to use ECC DRAM, then an alternative would be to use enterprise-grade DRAM, with higher long-term reliability.
Table 1. Compares the types of available flash media & various program/erase (P/E) cycles
3. Performance
In early IIoT implementations, systems were sometimes set to record-only, with data crunched and analysed later. But more recent IIoT systems, such as those using SMART Wireless Computing’s Qualcomm Snapdragon™ processor-based system-on-modules (SoM) and single-board-computers (SBC), often use fast memory and fast storage, where analysis is performed by the device in real-time. IIoT systems used for earthquake and tidal wave alert monitoring, for example, require this high performance. Some products produced by IIoT system suppliers use Low-Power DDR3 (LPDDR3) and LPDDR4 memory for lowest-power operation at maximum performance levels. The CPU, memory and storage media should be chosen together to align with the application. Today’s data requirements have become more intensive, as obtaining real-time data has become a top concern for making business-critical decisions.
4. Cost
IIoT hardware costs can be dramatically affected by what types of memory and storage are used. The cost of memory and storage used in IIoT systems can account for between 5% and 40% of the total hardware cost. A complex industrial automation system which uses machine learning algorithms to identify key metrics and automatically adjust its settings in real-time based on predetermined preferences, for example, will be at the higher end of the scale. This is due to the fact that it requires more memory and storage versus a custom electrocardiogram (ECG) tracker which analyses the electrical activity of a human heart and visualises sensor data via a mobile app, for instance.
5. Power consumption
Lowering system power consumption is beneficial for reducing energy costs and thermal output, while increasing system reliability. A low thermal output system enables a compact fan-less design, which is a common requirement for ruggedised IIoT applications. A popular DRAM solution is to use LPDDR4. LPDDR4 is used for mobile devices and leveraged for IIoT. To save on energy, LPDDR4 DRAMs lower the nominal operating voltage to 1.1V. The LPDDR4 standard supports an improved power saving low frequency mode, which can bring the clock speed down for further battery savings when performing simple background tasks. For storage, two common solutions are used: UFS 2.0 or MicroSD cards. UFS 2.0 is a soldered down flash embedded MultiMedia Controller (eMMC) part, which is fast at both reading and writing to storage, and can complete each task simultaneously. These are widely used in mobile applications to extend battery life. The energy consumption associated with UFS 2.0 is approximately one milliwatt (mW) of power consumed during operations, and below 0.5mW when in the standby condition.
SMART Modular Technologies_Memory & storage for IIoT applications image
6. Longevity
The average expected deployment of IIoT system hardware is 7 to 10 years, and systems are often located in outdoor or other challenging environments, where swapping out equipment with next-generation systems is costly and unrealistic. Meanwhile, typical memory and storage process node migration timeframe is from 18 to 24 months, and technology migration from DDR3 to DDR4 or MLC NAND Flash to TLC NAND Flash is required every 3 to 5 years. Needless to say, there is a mismatch. IIoT systems require long-term support and serviceability. Ideally, support for memory and storage components should be aligned as close as possible with the service life of the IIoT equipment. Supply chain support requirements need to be factored into the architecture definition stage for IIoT systems.
7. Scalability
Like longevity, scalability is a critical element for IIoT systems, which must often scale to support tens of thousands of controllers, robots, machinery and other purpose-built applications. This may mean increasing the density of the memory and storage components used in the systems, and/or being able to upgrade existing systems to use higher performing products. DRAM and flash technologies are continually increasing in density. However, video resolution and colour depth, for example, have also been increasing and require correspondingly larger storage (SSD on device, and SD for portability and scalability).
8. Ruggedness
IIoT systems are often deployed in operating environments that include exposure to vibration, moisture and poor air quality. In these conditions, memory and storage components must be fully protected to avoid field failures. The most common preventative measures to these conditions include the process of conformal coating and underfill, as well as the selection of specialised components that are resistant to sulphur dioxide contamination. Field failures under these operating conditions usually don’t appear until the third year of operation. Testing and qualification procedures for IIoT systems should include exposure to these conditions. As the potential of Industry 4.0 begins to emerge, predictive maintenance is one technique being tested. Predictive maintenance techniques are designed to help determine the condition of in-service equipment in order to estimate when maintenance should be performed. This approach promises cost savings over routine or time-based preventative maintenance, because tasks are performed only as required. The ability to transmit real-time information from a factory floor to the cloud in an IIoT application has the potential to ensure production continues uninterrupted.
Figure 2. An example of a DDR4 MIP (Module-in-a-Package™) by SMART Modular Technologies
9. Temperature
The commercial temperature operatingrange for memory and storage technologies is 0°C to +70°C. Excessive temperature is one of the most common causes of memory and storage failures. Using memory and storage products that are rated to industrial temperatures (-40°C to +85°), such as DDR4 ECC SO-DIMMs or a DDR4 MIP (Module-in-a-Package™) from SMART Modular Technologies, can dramatically increase the reliability of the system.
10. Security
Data integrity and IIoT edge security has become increasingly critical. Protecting the IIoT edge system hardware from cyber-attack should be included in system design considerations. Both memory and storage products used in IIoT system can accommodate higher levels of security. For SSDs, Trusted Computing Group (TCG) Opal is the standard that defines authentication on a drive that is stronger and more feature rich than the standard 256-bit password provided by ATA – and when combined with 256-bit encryption, the drive is more secure than ever. Alternatively, the Federal Information Standard (FIPS) Publication 140-2 (FIBS PUB 140-2) is a US standard for protecting government secrets, requiring specific levels of Advanced Encryption Standard (AES), authentication (such as TCG Opal), tamper protection and electronic emissions control. IIoT edge systems may increasingly require one of these security standards. For DRAM, a relatively new security option for DRAM modules is available: it uses the Register Clock Driver (RCD) on registered DIMMs to detect and record unauthorised commands, and addresses access from the CPU to the DRAM; up to 16 customer-defined security rules can be added for additional protection.
Summary
Selecting the right memory and storage components to support IIoT systems will have increasing importance as Industry 4.0 continues to expand from common applications used in manufacturing and industrial automation to transport, energy, agriculture and smart cities. The movement, storage and real-time analysis of IIoT data is critical to the success of IIoT applications in the field, and electronics design engineers should carefully consider how memory and storage technologies can be integrated within IIoT hardware to offer increased performance, endurance, reliability, longevity and scalability, while helping to reduce hardware costs, lower power consumption and enhance security.
Contact Details and Archive...
