Choosing solid-state storage solutions for single board computers: Designers nightmare?
As PC processor performance has increased, so have the requirements for data bus and interface between storage devices and the processor. System architects and engineers are constantly challenged with the quickly evolving technologies of processors and chipsets as well as the plethora of solid-state storage devices. Paul discusses the designer’s dilemma of searching for the appropriate solid state storage solution while managing the technology evolution dynamics in both the processor and the flash markets.
The task of designing a new stand-alone embedded board or processor mezzanine - for each new generation processor or implementing faster buses or aligning storage technologies to processors and interfaces is clearly monumental in decision making. Such tasks would be an added burden on the entire engineering organization, which is already thinly stretched.
Embedded system architects across different markets from AdvancedTCA Single Board Computers (SBCs) and COM Express to PC/104 and other small form factors share a common high-level perspective. Typically they follow the edicts of continually improving form, fit, and function. Common goals include minimizing current and future design risks, lowering product life-cycle costs through module scalability and interchangeability, and planning smooth transition from legacy to legacy-free interfaces while allowing room for growth.
Embedded system architects roll up their sleeves and deal with the practical issues at hand. They look in their toolbox and begin the ritual of reconciliation between what they want and the available technology. An optimal solution provides rugged, reliable solid-state storage and uses a rugged industry standard socket and multifunction components to save board space, at the same time increasing memory capacity, speed, and performance. Embedded system architects must also ask if the technology and interface can be sustained for the long term.
In saving costs and time, the ecosystem for SBC, COM Express, small form factors, and AdvancedTCA markets depend upon consumer off-the-shelf storage technologies such as Compact Flash (CF) with Intelligent Drive Electronics (IDE) and Flash with Universal Serial Bus (USB) drives. Unfortunately, IDE has existed in substantially the same form since 1989. Firmware for CF has dramatically improved, but the advantages were mostly made for consumer digital applications and not for the rugged applications for those users in the ecosystems previously mentioned. In addition, engineers must contend with technology changes and end-of-life issues. Legacy interfaces including Peripheral Component Interconnect Extended (PCI-X), IDE, and PATA can no longer satisfy the requirements of current applications in medical imaging, industrial automation, ruggedized portable electronics, military, gaming, and retail point-of-sale applications.
These applications need high-speed storage that offers small form factor compactness, along with long service life and scalability. Bottlenecking higher speed on stand-alone embedded processor boards or processor mezzanines are data throughput and boot time. Serial Advanced Technology Architecture (SATA) is the latest generation of interface offered by chip makers for connecting storage devices and other peripherals to the CPU. Yet with this latest technology, design issues still exist with SATA assembly, capacity, reliability, and limitations.
To help with this technology gap, chipset manufacturers have added SATA on the board as an option, but design engineers are not able to take advantage of the SATA technology at the chipset level because there has been a profusion of incompatible approaches and standards. For instance, no form factor has been available that allows designers to assemble storage efficiently and sensibly. The form factors available for storage were the bulky Hard Disk Drive (HDD), the large metal encased Solid-State Drive (SSD), Flash on PCB with USB interface, or DiskOnChip.
Another dreadful thought lurks in the back of the designer's mind: A single HDD on the SBC is a high-probability single point of failure. Most system architects do not have a clear sense that the consumer off-the-shelf technologies will be supported, and life-cycle management is a heart attack away from being dead on arrival. In addition, designers have been challenged with product ruggedness as well as thermal management issues (especially with onboard disks) in relation to the tall profile disks that restrict airflow and power consumption.
What happens if designers need SATA storage? What do they do? The only option for the design engineer is to run the pinout to a connector and leave it there for the end user or another system designer to solve the problem. SATA's four pins make HDD and SSD the available choices. Obviously, these options rule out CF and USB. In many cases neither HDD nor SSD is a viable solution because both form factors are too big. Unfortunately, there is not a good SATA option for these flash storage choices because of their form factors. In some cases the SATA drive eliminates the height issue, but lacks the robustness required for high availability platforms.
Four logical options
Let's review the four logical options currently being considered by design engineers: CompactFlash, USB, Disk-on-Chip (DOC), and SATA module; and look at the pros and cons of each. This will provide a starting point for matching applications in SBC, COM Express, PC/104, and AdvancedTCA based upon performance.
CF is a storage device (See Figure 1, courtesy of Performance Technologies and Virtium Technology) used in portable electronic devices and typically uses flash memory in a standardized metal case enclosure. With its design flexibility, the CF card is able to support a broad range of customers from industrial to embedded applications.
USB flash disk module
The USB flash disk module consists of nonvolatile, solid-state storage NAND-type flash memory. USB modules like the one in Figure 2 (courtesy of Kontron) are well suited for embedded applications and are available in capacities up to 8 GB. The modules are targeted for the embedded market, such as gaming, workstations, networking equipment, and industrial PCs.
The SATA module is a dual function memory module that combines SSD and Synchronous Dynamic RAM (SDRAM) technologies into a single module that conforms to the industry-accepted SODIMM form factor. The SATA Module on SODIMM in Figure 3 is courtesy of Virtium Technology, Inc.
The key points of the SATA module enable a long-term sustainable solution directly on the CPU board and allow engineers to design with the latest generation of storage interface offered by chipset manufacturers. Additionally, the SATA module can solve industry-wide uncertainty regarding current and future generation storage form factors and interfaces.
Another technology that should be included in a general discussion is DiskOnChip. It is a flash storage device like a hard disk drive but it is on a very small chip. One of its attractive features is flexibility because it is a solution for fast and reliable data storage. However, its poor thermal properties and difficulty in assembly do not make DOC a viable option for rugged applications.
Table 1 summarizes these four storage products with their technology performance, form-fit-function, scalability, interchangeability, transition from legacy to legacy-free interface, headroom for growth, and life-cycle management.
Although not obviously apparent, the logical technology of choice is the SATA interface SSD sharing the SODIMM socket with the main memory. This juxtaposed solution is the genesis of Virtium's SSDDR SODIMM that is a dual function module. Figure 4 is a SSDDR SATA module, courtesy of Virtium Technology, Inc. The "SSD" represents the Solid-State Drive and "DDR" is for Double Data Rate. Voila! SSDDR on a SODIMM socket.
SSDDR products make long-term sustainable storage solutions possible by allowing embedded system architects to design with the latest generation of high-performance storage interface offered by chipset manufacturers. Additionally the SSDDR SODIMM's main memory and storage technologies conform to accepted industry standards. SSDDR SODIMM can be easily adopted and implemented in the system, as both the storage and memory functions share the existing JEDEC Standard socket. System architects gain flexibility in their designs with SSDDR SODIMM as they can use its dual functionality or go back to the single function DDR SODIMM as the main memory on board. SSDDR provides an acceptable alternative to SSD and rotating disk drives in some applications and can even improve the overall reliability of the platform.
There seems to be no limit to the number of applications for which SSDDR can be considered - industrial SSD, SATA storage module, and embedded SSD board design. SSDDR has the demonstrated ability to meet AdvancedTCA storage system goals by increasing storage performance, decreasing the thermal impact, increasing reliability, and meeting cost targets.
Economic cost is the driving factor behind design solutions and methods for most embedded applications. In the past, SSD as storage technology was commonly known to be five to seven times more expensive than consumer HDD, CF, and USB. However designers have learned that headroom for product growth plus long term sustainable solution and product support triumph and are the best measures of investment. Designers are finding that the upfront investment for a long-term solution is less painful than the costs of succumbing to consumer, off-the-shelf technologies that suddenly can go out of existence.
Paul Dinh is the Director of Technical Marketing at Virtium Technology, Inc. During the past 20 years, Paul's professional work experiences have encompassed leadership roles in both the technical and business arenas. He has an extensive background in strategic marketing management, strategic account management, product management, and outside sales. His strong technical skills positioned him to be a member of the technical staff as well as involved in applications engineering and product engineering positions. He has experience in diverse markets such as aerospace, biotech, instrumentation, pharmaceutical, semiconductor, and government and military markets. Paul holds a BS degree in Physics from the University of California, Los Angeles (UCLA) and an MBA from the University of California, Irvine (UCI).