S.M.A.R.T., which stands for Self-Monitoring, Analysis, and Reporting Technology, is a set of drive health indicators for monitoring the quality and performance of disk operations, analyzing drive issues, and reporting the condition of the disk drive to the host.

What are SMART Attributes? 

SMART attributes include information observed by the drive over the course of its operation, such as temperature, power-on hours, and reallocated sector count. Each SMART attribute may come with six pieces of data: a raw value, a normalized value, a threshold, a worst value, a pre-fail/advisory flag, and an online data collection flag.

The raw value is simply the raw value of the SMART attribute being measured. The normalized value is the SMART attribute scaled to some value between 1 and 253, where values greater than 100 are typically good, and higher values are typically better. The threshold is the normalized value below which a SMART attribute will indicate that it has been “tripped.” A SMART trip can be thought of as an alarm sent from the drive to the host that requires an immediate response. When a SMART trip occurs, users are advised to replace their drive. The pre-fail/advisory flag is set to true if the manufacturer deems the attribute to be an important indicator of drive failure. The online data collection flag is set to true if the attribute is collected while the drive is in normal operation.

Issues with Standardization

SMART Attributes have value when it comes to predicting drive failures. Many papers have been published where SMART Attributes have been used to train AI algorithms that predict such failures. However, one drawback of SMART is that it is not always standardized, especially for predicting drive failure. Its interpretation depends on the manufacturer or vendor specifications, and its adoption is not required for any industry certification.

However, other drive health indicators, such as those obtained according to industry specifications such as the Information Technology ATA Command Set (1) (for SATA drives), the NVMe Express specification (2) (for NVMe drives), and the Information Technology SCSI Primary Commands specification (3) (for SCSI/SAS drives) are much more widely standardized compared to SMART attributes, and may provide more consistent measures of drive health as a result.

Furthermore, many SMART Attributes have equivalent metrics that can be found using the command sets listed in the aforementioned specifications. For example, power-on-hours can be found in SMART 9, but can also be found using a SATA command that pulls up certain drive logs. Analogous commands exist for NMVe and SCSI drives.

Ultimately, the choice of which set of health indicators to study comes down to the availability of the health indicator among the drives one wishes to study. That is, some health indicators may only be available as either SMART Attributes or non-SMART attributes on some drive models.

In a data dependent world, the ULINK DA Drive Analyzer AI algorithms disrupt drive failure prediction with 7-8 times more effectiveness than traditional systems.

 

1) Weber, R. (2013). Information Technology–ATA/ATAPI Command Set-3 (ACS-3). Working Draft Project American National Standard, 13.
2) https://nvmexpress.org/specifications/
3) Weber, R. O. (2005). Information Technology—SCSI Primary Commands—4 (SPC-4). revision 23. Technical report, INCITS Technical Committee T10/1416-D.

 

QNAP and ULINK Release DA Drive Analyzer, AI-powered Drive Failure Prediction Tool for NAS 

Photo Credit: bluebeat76