Enhancing Storage Efficiencies from Greater Understanding of SSD Application Classes

by Scott Phillips, Virtium

The simple dollar per gigabyte metric no longer applies in evaluating storage solutions now that embedded systems computing has migrated from desktops and laptops to mobile devices and cloud storage. Developers have started to use a more sophisticated review process based on their specific application and its related data-type usage model. NAND flash technology employed by SSDs has also freed developers from the physical and mechanical spinning platter limitations of hard drives. In turn, SSD manufacturers have capitalized on the technology’s inherent advantages to optimize their solutions based on the application at hand.

With the SSD industry continuing to explode, manufacturers are looking for ways to competitively position their products through technology differentiation and application-specific branding. This has caused several application classes to emerge. These classes are referred to as client, enterprise, data center and embedded SSDs. For the most part, SSDs are made up of the same components: a controller (ASIC or FPGA), NAND flash (other technologies are on the horizon but none are yet as commercially viable), and possibly DRAM. These components are either integrated into a solder-down multichip package or are combined with other passives and mounted onto a printed circuit board of some type.

So what characteristics differentiate SSDs built for each application class? The common answer from most SSD industry professionals is to define how the product is built as opposed to what the product does. They typically get into deep discussions about MLC vs. SLC vs. TLC (and now vs. 3D), write amplification optimization, read disturb mitigation, voltage threshold shifting and myriad other “secret sauces.” In the end, do any of these parameters really matter to system developers? Probably not. What is most important is that SSDs address developers’ needs by meeting the objectives of the applications’ usage models. Developers are keenly focused on finding the right storage solutions that fit their budget and application specs. That is why it is important to highlight the key external metrics of client, enterprise, data center and industrial SSDs without being overly concerned with the underlying technology and how these metrics are accomplished.

SSD Application Usage Models

To fully understand the reasons behind the different SSD application classes, let’s look at an overview of the applications for which they are designed.

There are many well-known use cases and metrics associated with client applications that include desktops, notebooks and now ultrabooks, tablets and smartphones. In these applications, SSDs are used for storing operating systems and user data that is generated or downloaded by a single person. Performance is largely subjective based on individual needs with the most desired features being instant-on and application response time. Client SSDs are typically optimized for read speed. Write speed doesn’t matter as much, and there is quite a bit of downtime associated with client applications. This downtime is enough for the SSD to take care of any background task, such as flash management, that will help it achieve higher performance, greater reliability or longer endurance.

Enterprise class SSDs were originally developed to replace racks of short-stroked enterprise class hard drives. In recent years, SAS has become the interface of choice for storing higher-reliability, mission-critical enterprise data that dictated the development of enterprise class SSDs, which use the same interface. SAS, with its dual port modes, DIF and other data integrity enhancements, offers the benefit of higher reliability than SATA. However, the performance capabilities of SSDs, which quickly highlighted the traditional hard drive interfaces as a major bottleneck, so impacted the enterprise that even more performance was desired. Enter PCIe as the high-performance interface of choice for today’s most demanding enterprise applications. Today, enterprise SSDs fall into three basic categories as shown in Table 1.

Table 1
The basic categories of today’s solid state storage devices.

Table 1
The basic categories of today’s solid state storage devices.
Data Center SSDs are designed as the main storage building block for application-specific servers and appliances. Driving this application class are multiple Internet search and social media sites. SSDs for the data center are generally 6 Gbit/s SATA SSDs in capacities of 120 Gbyte and higher. SATA is typically chosen because it is a well-known, straightforward interface that is highly compatible and generally the most cost-efficient compared to SAS and PCIe. It is no wonder then that the SSD companies that stand out are those that manufacture NAND flash. For purposes of this analysis; data center SSDs are positioned for a lower cost per gigabyte while maintaining adequate IOPS and low latencies, and generally feature read/write speeds around 500 Mbyte/s and IOPS in the 60K+ range.

The utilization of SSDs in industrial and embedded systems comes into clear focus with their deployment mainly in equipment that supports the infrastructure. Examples of infrastructure applications include networking and communication industry routers, switches and base stations; enterprise network security and monitoring devices; medical and gaming equipment; factory automation systems and digital signage…and the list goes on.

Compared to the well-known usage models for client and enterprise SSDs, infrastructure SSD applications are highly fragmented, making them more difficult to segment into a particular application class. As noted previously, client SSDs are employed for read use cases and enterprise SSDs are used to support write-intensive workloads. Infrastructure SSDs, on the other hand, need to support a wide range of mixed function workloads. Two opposite examples are casino gaming and radio base station applications. Casino gaming SSDs might only be written to once and then be write protected, but are read from as games are played. Base station SSDs may need to be continuously written with cell phone traffic log information. Infrastructure equipment data patterns can range from 99% read/ 1% write to just the opposite and can encompass every scenario in between.

Infrastructure applications are also considered mission-critical and must be designed for 24/7 operation—many times in harsh, extended temperature environments ranging from -40° to 85°C and higher. Embedded infrastructure-based SSDs are frequently characterized by smaller, lower power, lower capacity form factors such as Slim SATA, mSATA, CompactFlash or 10-pin eUSB whose applications typically require capacities less than 100 Gbyte—with a large number of Linux and real time OS-based systems requiring less than 4 Gbyte (Figure 1).

Figure 1
StorFly SATA SSDs are available in seven different industry standard form factors – 2.5”, 1.8”, M.2 22×42, M.2 22×80, mSATA, Slim SATA and CFast – and in capacities ranging from 8 Gbyte to 480 Gbyte.

A common belief among infrastructure system developers is that industrial SSDs need to be built with SLC NAND, making them considerably more “expensive” than client SSDs. This is not necessarily true. While SLC is more expensive on a dollar per Gbyte basis, there are many applications where the lowest-cost 120 Gbyte client SSD is still more expensive than the “right” 8 Gbyte SLC infrastructure SSD on a dollar per unit basis. There are also numerous mission-critical scenarios where SLC-based SSDs are essential, so the expense is justified with greater endurance and reliability over a longer product lifecycle.

Embedded system developers are also concerned with the high cost of requalification. The reality today is that there may be three iterations of MLC for every one iteration of SLC, which necessitate a requal for every iteration. For high-capacity requirements, it may still be difficult to cost-justify using SLC, but as capacity needs decrease, the total cost of ownership (TCO) and performance arguments for SLC become more compelling (Figure 2).

Figure 2
Virtium offers some of the storage industry’s longest product lifecycles, which help minimize costly and disruptive requalifications. Its second generation SLC-based StorFly PE class products are guaranteed to not cause a requal for at least four years.
Rules Are Not Always Followed

After reviewing the diverse set of applications, it should be abundantly clear why SSD application classes are defined by usage model and their associated workload requirements rather than technology. These definitions provide a helpful set of guidelines in specifying SSDs. However, not all SSD suppliers follow these guidelines, and it is not mandatory to do so. At the moment, the JEDEC JC-64.8 SSD committee defines application classes only for client and enterprise SSDs in document JESD218. The workloads associated with these application classes are explained in JESD219.

Unfortunately, embedded system developers may not find a given SSD specification to be particularly useful or meaningful if it isn’t based on a set of common rules. JEDEC definitions are helpful in specifying client and enterprise SSDs, but they don’t cover all of the considerations for embedded SSDs. Therefore, it is important for designers to carefully review datasheets to fully understand the assumptions and conditions under which the products are specified (Figure 3).

Figure 3
Virtium’s StorFly SATA, TuffDrive USB and CompactFlash SSDs match the use case needs of a multitude of industrial embedded systems. StorFly SATA SSDs are not only available with SATA 6G interfaces, they can also be configured as SATA 3G or even SATA 1.5G devices, since many legacy systems were never designed to accommodate handshaking from SATA 6G.

Validating endurance for an infrastructure application class workload is an excellent example where designers may need to define many elements that include active use (power on) time and temperature, retention use (power off) time and temperature, and functional failure and uncorrectable bit error rate requirements. Adding to the challenge is that the metrics shown in Table 2 are all interrelated when it comes to endurance, and changes in assumptions for one parameter can lead to changes in another.

SSDs That Match Industrial Infrastructure Needs

Because of the broad and varying storage requirements of embedded and industrial systems infrastructure applications, OEMs today need to seek multiple options to match their individual system needs. While there are many suppliers of SSDs, not all are in tune with embedded infrastructure application developers’ unique requirements. These developers require specifically engineered SSDs that address unique challenges including power-down protection, 24/7 availability, reliable operation over a wide temperature range, low-power/low-heat, high endurance and long product lifecycles, to name a few.

As the analysis should have highlighted, total cost of ownership and improved storage efficiency benefits can be achieved in embedded system designs by taking the time to fully understand SSD application classes. This information can greatly assist OEMs in selecting the most optimal solution for their particular design. Although it can be challenging to find the right SSD to meet budget and application specifications, there are expert storage suppliers ready to take a larger role in serving the needs of this diverse market with extensive expertise in guiding OEMs to the optimal storage solution, and by delivering high-reliability, quality products that are proven to work over the course of the systems’ lifecycle goals.

Santa Margarita, CA.
(949) 888-2444