Making Sense of SSD SMART Stats

Over the past several years, folks have come to embrace the solid state drive (SSD) as their standard data storage device. It’s gotten to the point where people are breathlessly predicting the imminent death of the venerable hard drive. While we don’t see the demise of the hard drive happening any time soon, SSDs are here to stay and we want to share what we know about them. To that end, we’ve previously compared hard drives and SSDs as it relates to power, reliability, speed, price, and so on.  But, the one area we’ve left primarily unexplored for SSDs is SMART.

Both hard drives and SSDs populate SMART attributes, but given how different these drive types are, the information produced is quite different as well. For example, hard drives have sectors, while SSDs have pages and blocks. Let’s take a look at the common attributes of hard drives and SSDs, and then we’ll dig into the SSD SMART attributes we’ve found useful, interesting, or just weird.

Let’s Get SMARTed

For each SSD model, the drive manufacturer decides which SMART attributes to populate. Attributes are numbered from 1 to 255, with raw and normalized values for each attribute. Some SMART reference material will also list attributes in hexadecimal (HEX), for example, decimal 12 will also be shown as “HEX 0C.”  

At Backblaze, we have over a dozen different SSD models in service, and we pull daily SMART stats from each. To simplify the task at hand for the purposes of this blog post, we chose three SSD models, one each from Seagate, Western Digital, and Crucial, to show the similarities and differences between the models. All three are 250GB SSDs.

To that end, we have created a table of the SMART attributes used by each of those three drive models. You can download a PDF of the table, or jump to the end of this post to view the table.  Things to note about the table:

• Only 44 of the 255 available attributes are used by these SSDs. Most of the other attributes are exclusive to hard drives or not used at all.
• The attribute names and definitions were gathered from multiple sources which are referenced at the end of this post. The consistency of the names and definitions across all SSD manufacturers is, well, not as consistent as we would like.
• Of the 44 attributes listed in the table, the Seagate SSD (model: Seagate BarraCuda 120 SSD ZA250CM10003) uses 20, the Western Digital (model: WDC WDS250G2B0A) uses 25, and the Crucial (model: CT250MX500SSD1) uses 23.
• The SMART values listed for each SSD model are those that were recorded using the smartctl utility from the smartmontools package. 

One of the things you’ll notice as you examine the list of attributes is that there are several which have similar names, but are different attribute numbers. That is, different vendors use a different attribute for basically the same thing. This highlights a deficiency in SMART:  Participation is voluntary. While the vendors try to play nice with each other, who uses a given attribute for what purpose is subject to the whims, patience, and persistence of the many SSD manufacturers in the market today.

Often manufacturers have created their own SMART monitoring tools to use on their drives. As they add, change, and delete the SMART attributes they use, they update their tools. Drive agnostic tools such as smartctl, which we use, have to chase down updates that have occurred in each of the manufacturer’s homegrown SMART monitoring tools. There are other tools out there as well. DriveDX is another vendor-agnostic SSD monitoring tool, and here’s a link to their release notes page. They made 38 updates in release 1.10.0 (700) alone just to keep up with the drive manufacturers.

Making things more complicated, manufacturers differ widely in how they advertise the attributes and definitions they use. Kingston, for example, is very good about publishing a table of named SMART attributes and definitions for each of their drives, whereas similar information for Western Digital SSDs is difficult to find in the public domain. The net result is that agnostic SMART tools such as smartctl, DriveDx, and others have to work extra hard to keep up with new, updated, and deleted attributes.

Common Attributes

Of the 44 attributes we list in our table, only five are common for all three of the SSD models we are examining. Let’s start with the three of the common attributes that are also common to nearly every hard drive in production today.

• SMART 9: Power-On Hours. The count of hours in power-on state.
• SMART 12: Power Cycle Count. The number of times the disk is powered off and then powered back on. This is cumulative over the life of the drive.
• SMART 194: Temperature. The internal temperature of the drive. For some drive models, the normalized value ranges from 0 to 255, for other drive models the range is 0 to 100, and for others the normalized value is the same as the raw value. In all cases, the raw value is in degrees Celsius.

SSD Unique Common Attributes

These two attributes are specific to SSDs and are common to all three of the models we are examining.

• SMART 173: SSD Wear Leveling. Counts the maximum worst erase count on a single block.
• SMART 174: Unexpected Power Loss Count. The number of unclean (unexpected) shutdowns, like when you kick out the plug of your external drive. This value is cumulative over the life of the SSD. This attribute is a subset of the count for SMART 12 and with a little math you can get the number of normal shutdowns if that is interesting to you.

Not Much In Common

As noted, only five of the 44 SMART attributes are common between our three SSD models. This lack of commonality, 11%, seemed low to us, and we wondered what the commonality was between the SMART attributes on the hard drive models we use. We reviewed the SMART attributes for three 14TB hard drive models in our drive stats data set, one model each from Seagate, Western Digital, and Toshiba. We found that 42% of the SMART attributes were common between the three models. That’s nearly four times more than the SSD commonality, but admittedly less than we thought. 

Useful Attributes

For the purpose at hand, we’ll define a useful attribute as something that clearly indicates the health of the SSD. That led us to focus on two concepts: Lifetime remaining (or used) percentage, and logical block addressing (LBA) read/write counts. Let’s take a look at how each of the drive models reports on these attributes.

Lifetime Percentage

SMART 169: Remaining Lifetime Percentage (Western Digital)

This attribute measures the approximate life left from a combination of program-erase cycles and available reserve blocks of the device. A brand new SSD will report a value of “100” for the Normalized value and decrease down to “0” as the drive is used.

SMART 202: Percentage of Lifetime Used (Crucial)

This attribute measures how much of the drive’s projected lifetime has been used at any point in time. For a brand new drive, the attribute will report “0”, and when its specified lifetime has been reached, it will show “100,” reporting that 100 percent of the lifetime has been used. 

SMART 231: Life Left (Seagate)

This attribute indicates the approximate SSD life left, in terms of program/erase cycles or available reserved blocks. A brand new SSD has a normalized value of “100” and decreases from there with a threshold value at “10” indicating a need for replacement. A value of “0” may mean that the drive is operating in read-only mode.

All three use program/erase cycles (SMART 232) and available reserved blocks (SMART 170) to compute their percentages, although as is seen, SMART 202 counts up, while the other two count down. Lifetime, as defined here, is relative. That is you could be at 50% lifetime after six months or six years depending on the SSD usage. 

LBAs Written/Read

In an SSD, data is written to and read from a page, also known as a NAND page. A group of pages forms a block. The LBA written/read count is just that, a count of blocks written/read. Each time a block is written or read the respective SMART attribute counter increases by one. For example, if various pieces of data on the pages within a single block are read 10 times, it will increase the SMART counter by 10. 

SMART 241: LBAs Written (Seagate and Western Digital) 

Total count of LBAs written.

SMART 242: LBAs Read (Seagate and Western Digital)  

Total count of LBAs read.

SMART 246: Cumulative Host Sectors Written (Crucial)  

LBAs written due to a computer request. Note that the name of this attribute seems incorrect as it states sectors versus blocks. 

Crucial also counts NAND pages written due to a computer request (SMART 247) and NAND pages written due to a background operation such as garbage collection (SMART 248). Crucial does not seem to have a SMART attribute for total count of LBAs read. Nor does it seem to record LBAs written for background operations.

Interesting Attributes

Below we’ve gathered several SSD SMART attributes we found interesting and one could argue potentially useful. In no particular order, let’s take a look.

SMART 230: Drive Life Protection Status (Western Digital)

This attribute indicates whether the SSD’s usage trajectory is outpacing the expected life curve. This attribute implies a couple of interesting things. First, there is a usage trajectory calculation and value. This could be SMART 169 noted previously. Second, there is a defined expected life. We assume that the expected life curve is fixed for a given SSD model and perhaps uses the warranty period as its zero date, but we’re only guessing here. 

SMART 210: RAIN Successful Recovery Page Count (Crucial)

Redundant Array of Independent NAND (RAIN) is similar to gaining data redundancy using RAID in a drive array, except RAIN redundancy is accomplished within the drive, i.e., all the data written to this SSD is made redundant on the SSD itself. This redundancy is not free and either consumes some of disk space from the total space specified (250GB in this case), or uses additional space not counted in the total. Either way, this is a really cool feature and allows for data to be recovered transparently to the user even when initially it couldn’t be read due to a bad page or block.

SMART 232: Endurance Remaining (Seagate and Western Digital)

The number of physical erase cycles completed on the SSD as a percentage of the maximum physical erase cycles the drive is designed to endure. At first look, this seems similar to SMART 231 (Life Left), but this attribute does not consider available reserved blocks as part of its calculus. Still, this attribute could be a harbinger of what’s to come, as erasing SSD blocks at an accelerated rate often leads to having to utilize available reserved blocks downstream as the SSD cells wear out.

SMART 233: Media Wearout Indicator (Seagate and Western Digital)

Similar to SMART 232 (but without the math) as this attribute records the count of the actual NAND erase cycles. The normalized value starts at 100 for a new drive and decreases to a minimum of 1. As it decreases, the NAND erase cycles count (raw value) increases from 0 to the maximum-rated number of cycles.

SMART 171: SSD Program Fail Count (Western Digital and Crucial) and SMART 172: SSD Erase Count Fail (Western Digital and Crucial)

Both of these attributes count their respective failures (Program Fail and Erase Count) from when the drive was deployed. As a drive ages, one would expect these counts to increase and eventually pass some threshold value which would indicate a problem. While this is helpful in determining the health of a drive, these attributes alone provide only a partial picture as they can miss a rapid acceleration of failures over a short period of time. 

Weird Things

There are a handful of attributes which seem odd based on our table and the attribute names and the definitions we have found. We’d like to point these out to start the conversation—If anyone can shed some light on these oddities, jump in the comments. Your input is much appreciated.

SMART 16: Total LBAs Read (Seagate)

There are two odd things here. First, the definition states that this attribute is only found on select Western Digital hard drive models—yet it was found in most of our Seagate SSDs. This could be a definition problem, but then there’s the second thing: Seagate SSDs record Total LBAs Read in attribute 242 (noted above). So, it seems it could also be an attribute name problem. 

SMART 17: Unknown (Seagate)

We could not find any information on SMART 17, except for the fact that our Seagate drives report on this attribute. 

SMART 196: Reallocation Event Count (Crucial), SMART 197: Current Pending Sector Count (Crucial), and SMART 198: Uncorrectable Sector Count (Crucial)

Our Crucial drives report values for these attributes, but this is another case where the names and definitions don’t make sense, as they are talking about sectors which are hard drive-specific.

SMART 206: Flying Height (Crucial)

Another attribute reported by our Crucial drives which makes no sense based on the name and definition. I think we can all agree that measuring the flying height of the cells within an SSD is not meaningful.

The questions around the Crucial reported attributes could be straightforward to answer as Crucial has their own free SMART monitoring software, Storage Executive. If you are using this software, we’d appreciate any info you can share on the Crucial names and definitions of these attributes.

Data Retention

Many of us have an external hard drive or two sitting on a shelf somewhere acting as a backup or perhaps even an archive of our data. Every so often, we take out one of those drives, plug it in, and hope it spins up. This can go on for years. 

Can SSDs be used for offline data storage, and if so how long can they safely remain unplugged? It’s a good question and one that has been debated many times over the years with time frames ranging from a few weeks to several years. The current thinking is that when an SSD is new, it can safely store your data without power for a year or so, but as the drive wears out the data retention period begins to diminish. 

This begs the question: How worn out is your SSD? For Crucial SSDs, the answer is SMART 202: Percentage Lifetime Used. We discussed this attribute earlier in relation to drive life, but it also plays a role in data retention when the drive is unpowered. Using the normalized value, Crucial estimates the following: 

• “0” indicates that the drive can be stored unpowered for up to one year.
• “50” indicates that the drive can be stored unpowered for up to six months.
• “100” indicates that the drive can be stored unpowered for up to one month. 
• Anything above “100” and your data is at risk when the SSD is powered off.

In theory, you should be able to use the SMART 231: Life Left (Seagate) or SMART 169: Remaining Lifetime Percentage (Western Digital) to perform the same analysis as was done above with SMART 202 and the Crucial SSD model. Remember that these two attributes (231 and 169) count downward, that is “100” is good and “0” is bad. All that said, this is just a theory, as we’ve found no documentation this is actually the case (but it does seem to make sense).

SMART Could Be Even SMARTer

It’s great that SSD manufacturers are using SMART attributes to record relevant information about the status and health of their drive models. It’s also great that many manufacturers also provide software that monitors these SMART stats and provides the user feedback. All is wonderful when you are buying all your SSDs from the same manufacturer. But that’s just not the reality for most IT shops who are managing servers, networking gear, and so on from different vendors. It is also not the reality when it comes to running a cloud storage company. 

Having accurate, up-to-date, vendor agnostic SSD monitoring tools is important to many organizations as part of their ability to cost effectively manage their systems and keep them healthy. Having to use a multitude of different tools to monitor SSDs doesn’t benefit anyone. Maybe it’s time we take SMART for SSDs beyond voluntary and look to standardize the attributes and their names and definitions across the board for all SSD manufacturers.

Sources

Multiple sources were consulted in researching this post, they are listed below. We may have missed one or two sources, and we apologize in advance if we did.

• https://en.wikipedia.org/wiki/Self-Monitoring,_Analysis_and_Reporting_Technology
• https://en.wikipedia.org/wiki/Solid-state_drive
• https://media.kingston.com/support/downloads/KC600-SMART-attribute.pdf
• https://media.kingston.com/support/downloads/MKP_521_Phison_SMART_attribute.pdf
• https://media.kingston.com/support/downloads/MKP_306_SMART_attribute.pdf
• https://www.cropel.com/library/smart-attribute-list.aspx
• https://www.crucial.com/support/articles-faq-ssd/ssds-and-smart-data
• https://www.micromat.com/product_manuals/drive_scope_manual_01.pdf
• https://www.recoverhdd.com/blog/smart-data-for-ssd-drive.html

We only used sources which are available to us without purchasing something. That is, we didn’t buy agnostic monitoring applications or purchase a specific manufacturer’s SSD to have something to use their free monitoring application on. We took our Drive Stats data and then, just like you, we ventured into the internet to search out SSD SMART attribute information that was publicly available.

SMART Attributes Table

The following table contains the SMART attributes for the three SSD models listed. These attributes are collected by the smartctl utility within the smartmon toolset.