User Manual

Constellation ES.3 SAS Product Manual, Rev. F 19
5.2.3 S.M.A.R.T.
S.M.A.R.T. is an acronym for Self-Monitoring Analysis and Reporting Technology. This technology is intended to recognize conditions
that indicate imminent drive failure and is designed to provide sufficient warning of a failure to allow users to back up the data
before an actual failure occurs.
.
Each monitored attribute has been selected to monitor a specific set of failure conditions in the operating performance of the drive
and the thresholds are optimized to minimize “false” and “failed” predictions.
Controlling S.M.A.R.T.
The operating mode of S.M.A.R.T. is controlled by the DEXCPT and PERF bits on the Informational Exceptions Control mode page
(1Ch). Use the DEXCPT bit to enable or disable the S.M.A.R.T. feature. Setting the DEXCPT bit disables all S.M.A.R.T. functions. When
enabled, S.M.A.R.T. collects on-line data as the drive performs normal read and write operations. When the PERF bit is set, the drive is
considered to be in “On-line Mode Only” and will not perform off-line functions.
Users can measure off-line attributes and force the drive to save the data by using the Rezero Unit command. Forcing S.M.A.R.T.
resets the timer so that the next scheduled interrupt is in one hour.
Users can interrogate the drive through the host to determine the time remaining before the next scheduled measurement and data
logging process occurs. To accomplish this, issue a Log Sense command to log page 0x3E. This allows users to control when S.M.A.R.T.
interruptions occur. Forcing S.M.A.R.T. with the RTZ command resets the timer.
Performance impact
S.M.A.R.T. attribute data is saved to the disk so that the events that caused a predictive failure can be recreated. The drive measures
and saves parameters once every one hour subject to an idle period on the drive interfaces. The process of measuring off-line
attribute data and saving data to the disk is interruptable. The maximum on-line only processing delay is summarized below:
Reporting control
Reporting is controlled by the MRIE bits in the Informational Exceptions Control mode page (1Ch). An example, if the MRIE is set to
one, the firmware will issue to the host an 01-5D00 sense code. The FRU field contains the type of predictive failure that occurred. The
error code is preserved through bus resets and power cycles.
Determining rate
S.M.A.R.T. monitors the rate at which errors occur and signals a predictive failure if the rate of degraded errors increases to an
unacceptable level. To determine rate, error events are logged and compared to the number of total operations for a given attribute.
The interval defines the number of operations over which to measure the rate. The counter that keeps track of the current number of
operations is referred to as the Interval Counter.
S.M.A.R.T. measures error rates. All errors for each monitored attribute are recorded. A counter keeps track of the number of errors for
the current interval. This counter is referred to as the Failure Counter.
Error rate is the number of errors per operation. The algorithm that S.M.A.R.T. uses to record rates of error is to set thresholds for the
number of errors and their interval. If the number of errors exceeds the threshold before the interval expires, the error rate is
considered to be unacceptable. If the number of errors does not exceed the threshold before the interval expires, the error rate is
considered to be acceptable. In either case, the interval and failure counters are reset and the process starts over.
Note
The drive’s firmware monitors specific attributes for degradation
over time but can’t predict instantaneous drive failures.
Maximum processing delay
Fully-enabled delay DEXCPT = 0
S.M.A.R.T. delay times 75 ms