Design Article
Comment
KGround
Whichever technology of SSD you choose it is important to note that one can't ...
Specify the right SSD for industrial systems
Roger Griesemer, Swissbit AG
2/6/2012 3:21 PM EST
Until recently, high entry-level costs often discouraged the use of industrial solid-state drives (SSDs), but thanks to ongoing price optimization they are now becoming more and more popular as storage media. There are several key factors to take into consideration when deciding on the purchase of SSD storage media; however, varying storage applications will dictate the use of different memory solutions.
In the last few years it has become ever more common to use SSDs rather than traditional hard drives (hard disk drives, HDDs) in a growing range of applications. They supersede traditional storage media in many respects, predominantly in industrial markets. In particular, small SSD form factors such as CompactFlash or SD cards have already proven themselves in industrial environments for many years.
In general, SSDs are made from the following key components: a printed circuit board, a controller, firmware, memory (usually based on NAND flash technology), and optional cache. With NAND flash memory devices, individual storage cells pass into different states, thus storing information.
As this is done purely electronically, rather than with the use of magnetic rotating platters and read-and-write heads as in traditional HDDs, the access times of SSDs onto the entire memory area are much faster. SSDs are seeing even more interest in ruggedized industrial applications since they work without moving parts.
In comparison to HDDs, SSDs are less noisy and more reliable, energy efficient, and shock and vibration proof. Add to this a longer operational life span—nearly 10 years as compared to only a few years with HDDs—and their only disadvantage today is their increased price per density, which is continually decreasing. Two years ago, for example, a standard industrial SSD cost approximately $27 per gigabyte; today it is only $13 per gigabyte, and prices are still going down. This is because the structure size of flash devices decreased from 60 nm to around 20 to 40 nm and is likely to decrease even more. Although HDDs, costing in the area of $0.10 per gigabyte on average, are still clearly less expensive, SSDs can present the more favorable option when considered from the point of view of total-cost-of-ownership.
Many applications, such as those of the automation industry, require the use of SSDs for the advantages discussed above, most important of which is their many years of reliable service.
Three different SSD storage solutions
Using these novel memory solutions requires that designers take into consideration several aspects, however, since different SSDs have different functional properties. As a general rule, we differentiate between single-level cell (SLC), multi-level cell (MLC), and the novel triple-level cell (TLC) flash devices. SLC types store one bit per flash memory cell, MLC SSDs store two bits per cell, and TLCs store three bits per cell. MLC and TLC devices can thus be used to save much more information on the same device area than SLC versions because their storage density is increased. Because the die area is a significant cost factor in flash memory, MLC and TLC types are less expensive than their SLC counterparts. An MLC SSD costs $4 per gigabyte on average, whereas an SLC SSD costs about $13 per gigabyte.
Lowering cost involves a trade-off, however. It compromises the lifetime, reliability, speed, and endurance of the device. MLC memory therefore comes with a much higher failure rate, thus requiring much more extensive and elaborate error correction methodology.
MLC drives suffer from limited program erase cycles. Suppliers usually indicate 100,000 cycles for most SLC flash memories, but only 1,000 to 3,000 cycles for MLC versions. There are also significant differences in endurance. SLC solutions reliably store data for up to 10 years, whereas MLC solutions only store data for a maximum of one year. This clearly does not suffice for data logging or read-only applications as in boot media with static data for operating systems, for example.
These differences are technological. Take the following simplified example for clarification. A glass of water that can only be either full or empty simply and clearly shows its current state – just like an SLC storage cell that can be either 0 or 1. With four different fill levels like in MLC designs, however, it is more difficult to assess whether the glass is filled to 0%, 33%, 66% or 100%. The task becomes even more difficult with TLC components, which can present eight possible states (see figure 1).
In the last few years it has become ever more common to use SSDs rather than traditional hard drives (hard disk drives, HDDs) in a growing range of applications. They supersede traditional storage media in many respects, predominantly in industrial markets. In particular, small SSD form factors such as CompactFlash or SD cards have already proven themselves in industrial environments for many years.
In general, SSDs are made from the following key components: a printed circuit board, a controller, firmware, memory (usually based on NAND flash technology), and optional cache. With NAND flash memory devices, individual storage cells pass into different states, thus storing information.
As this is done purely electronically, rather than with the use of magnetic rotating platters and read-and-write heads as in traditional HDDs, the access times of SSDs onto the entire memory area are much faster. SSDs are seeing even more interest in ruggedized industrial applications since they work without moving parts.
In comparison to HDDs, SSDs are less noisy and more reliable, energy efficient, and shock and vibration proof. Add to this a longer operational life span—nearly 10 years as compared to only a few years with HDDs—and their only disadvantage today is their increased price per density, which is continually decreasing. Two years ago, for example, a standard industrial SSD cost approximately $27 per gigabyte; today it is only $13 per gigabyte, and prices are still going down. This is because the structure size of flash devices decreased from 60 nm to around 20 to 40 nm and is likely to decrease even more. Although HDDs, costing in the area of $0.10 per gigabyte on average, are still clearly less expensive, SSDs can present the more favorable option when considered from the point of view of total-cost-of-ownership.
Many applications, such as those of the automation industry, require the use of SSDs for the advantages discussed above, most important of which is their many years of reliable service.
Three different SSD storage solutions
Using these novel memory solutions requires that designers take into consideration several aspects, however, since different SSDs have different functional properties. As a general rule, we differentiate between single-level cell (SLC), multi-level cell (MLC), and the novel triple-level cell (TLC) flash devices. SLC types store one bit per flash memory cell, MLC SSDs store two bits per cell, and TLCs store three bits per cell. MLC and TLC devices can thus be used to save much more information on the same device area than SLC versions because their storage density is increased. Because the die area is a significant cost factor in flash memory, MLC and TLC types are less expensive than their SLC counterparts. An MLC SSD costs $4 per gigabyte on average, whereas an SLC SSD costs about $13 per gigabyte.
Lowering cost involves a trade-off, however. It compromises the lifetime, reliability, speed, and endurance of the device. MLC memory therefore comes with a much higher failure rate, thus requiring much more extensive and elaborate error correction methodology.
MLC drives suffer from limited program erase cycles. Suppliers usually indicate 100,000 cycles for most SLC flash memories, but only 1,000 to 3,000 cycles for MLC versions. There are also significant differences in endurance. SLC solutions reliably store data for up to 10 years, whereas MLC solutions only store data for a maximum of one year. This clearly does not suffice for data logging or read-only applications as in boot media with static data for operating systems, for example.
These differences are technological. Take the following simplified example for clarification. A glass of water that can only be either full or empty simply and clearly shows its current state – just like an SLC storage cell that can be either 0 or 1. With four different fill levels like in MLC designs, however, it is more difficult to assess whether the glass is filled to 0%, 33%, 66% or 100%. The task becomes even more difficult with TLC components, which can present eight possible states (see figure 1).
Figure 1: Single-level cell (SLC) memories come with two states (0 or 1), multi-level cell devices (MLC) come with four states, and triple-level cell (TLC) components come with a critical total of eight possible states to be assessed.
|
|
|
|
|
Navigate to related information
KGround
2/9/2012 9:56 AM EST
Whichever technology of SSD you choose it is important to note that one can't just drop in a SSD in place of a HDD. Even at 100,000 write cycles rating the SSD will be 'used up' in very short order (months in many cases) if the common (and default) Windows practice of using the disk for the Windows swap file is not disabled. This usually necessitates providing enough extra RAM to either provide space for a virtual drive or to make swap file use unnecessary. There is a place for SSDs in industrial applications, expecially where hard service like vibration kills HDDs quickly, but like any other component they must be applied sensibly and with regard to their limitations. Also, the comments on expected lifetimes of HDDs are interesting. It seems that as HDDs grew in capacity they aslo declined in longevity. We have some machines in the field using old 10 and 20Gb HDDs in datalogging operation with continuous use 24 hours a day that are over 20 years old. Now with new HDDs we recommend that customers plan on replacing them least every 5 years to avoid unexpected system failures, and sometimes they don't last even that long. It seems that PC components, even those used in 'industrial' computers are built for the disposable desktop market where the norm seems to be to replace the PC every few years. There is a definite but admittedly small market in industrial applications for smaller HDDs (or SSDs) with manufacturers' lifetime ratings meaasured in decades of continuous use instead of yesrs .
Sign in to Reply