SQFlash Intelligent SSD Self-Management

SQFlash Intelligent SSD Self-ManagementAuthor: Ares.ChengKeywords: SATA, SSD, S.M.A.R.T.To meet the increasing demand for higher data volumes, the unit storage density ofNAND flash has been continuously improved through substantial advancements inprocess technology that range from SLC to 3D NAND TLC. Yet, the downside ofthis improvement is reduced P/E cycles for high-density NAND flash and increasedrisk of data corruption. To facilitate monitoring of storage device status, the JEDECAssociation established relevant Self-Monitoring, Analysis, and ReportingTechnology (S.M.A.R.T.) specifications for SATA storage devices. However,because such instructions typically necessitate system technology-relatedknowledge, most users experience difficulty obtaining and utilizing such information,as well as performing adequate risk assessments and remote monitoring inaccordance with S.M.A.R.T. instructions.To reduce the risk of data corruption/loss and ensure data access, Advantechdeveloped its SQFlash Utility and WISE-PaaS/DeviceOn platform to help customersimplement S.M.A.R.T. specifications, predict product lifespans, and remotelymonitor system status in real-time using a simple graphical user interface. Thispaper provides a detailed description of the intelligent SSD self-managementsystem developed based on Advantech SQFlash.0www.advantech.com

Table of ContentIntelligent SSD Self-Management . 2SSD Information and Endurance Check . 2SSD Lifespan Prediction . 5General Prediction . 5Endurance Estimation . 6Remote Management . 7Predictive Maintenance Quality . 9Conclusion . 101www.advantech.com

Intelligent SSD Self-ManagementSSD Information and Endurance CheckS.M.A.R.T. information in the SATA specification contains up to 256information items. Some information is specific to traditional HDDs orSSD-related information, while some is considered manufacturer-defined orundefined information. To help users quickly understand the SSD statusand information, Advantech’s SQFlash Utility provides data for the 10 keyparameters listed in the Information column, as shown in Figure 1. The datais obtained by decoding and calculating S.M.A.R.T. attributes. Theremaining SSD endurance is presented in percentage using a pie chart.With this information, users can quickly understand the SSD status andestablish contingency measures accordingly.Figure 1. SSD Information and Endurance CheckAdditionally, the provision of easy access to S.M.A.R.T data via theSQFlash Utility allows users to obtain more advanced information regardingthe SSD status. As shown in Figure 2, many self-defined items are includedin the SQFlash S.M.A.R.T. data, such as 10h, 11h, 64h, A8h, AAh, ADh,AFh, DAh, E7h, EAh, and EBh.2www.advantech.com

Figure 2. SQFlash S.M.A.R.T. DataDATA ADDRESS (Byte)IDATTRIBUTE NAME10901h Raw Read Error Rate8765Uncorrectable ECC Count09h Power On HoursPower On Hours0Ch Power Cycle CountPower On/Off Counts0Eh Device Capacity00Device Capacity0Fh User Capacity00User Capacity00Total Available Spare Block11h Spare Blocks Remaining00Remaining Spare Block64h Total Erase Count00Total Erase Count10hInitial Spare BlocksAvailableA8h SATA PHY Error CountSATA PHY Error CountAAh Bad Block CountLater Bad00Early BadADh Erase Count00Avg. EraseMax Erase00AEhAFhUnexpected Power LossCountPower Failure ProtectionStatusUnexpected Power Loss CountVoltage Stabilizer Trigger Count3www.advantech.comGuaranteedDrive StatusFlush(0x00 Normal)

DATA ADDRESS (Byte)IDATTRIBUTE NAME1098765(0x01 Enable)C0hUnexpected Power LossCountC2h TemperatureCAh00Max. Temp.Percentage of SparesRemaining00000CountCurrent Temp.SSD Life0UsedCRC Error CountE7h SSD Life RemainingEBh Total NAND WrittenUnexpected Power LossMin. Temp.DAh CRC ErrorEAh Total NAND Read000000SSD Life LeftTotal NAND Read (Sector, 512B)Total NAND Written (Sector, 512B)F1h Total Host WriteHost Write (Sector, 512B)F2h Total Host ReadHost Read (Sector, 512B)Table 1. S.M.A.R.T. ID DefinitionsIDDescription of the SQFlash-Defined S.M.A.R.T. IDs10hInitial Spare Blocks Available: The spare block count when the SSD is newly made.11hSpare Blocks Remaining: Currently available spare blocks.64hTotal Erase Count: Sum of the erase count from all blocks.A8hSATA PHY Error Count: The PHY error count - such as data FIS CRC, code error,disparity error, command FIS CRC . etc. This value resets to zero when powered off.Bad Block Count: Block is a capacity unit of NAND flash. Bad blocks are damagedblocks that the SSD controller has marked as unusable.There are two types of bad blocks — early bad blocks and later bad blocks.AAhEarly bad blocks are the result of a manufacturing fault.Later bad blocks result from improper usage. If a block has too many ECCs, the SSDcontroller will mark the block as a later bad block. The erase count for the block will needto be checked. For SSDs nearing the end of their lifespan, later bad blocks are normal.Erase Count: This value is directly related to the SSD lifespan.There are two types of erase counts — the average erase count and max. erase count.The average erase count is the average of the erase count for all blocks.ADhThe max. erase count is the highest erase count of all blocks.The SSD endurance is dependent on the NAND flash type, as shown in Table 2.4www.advantech.com

IDDescription of the SQFlash-Defined S.M.A.R.T. IDsNAND Flash TypeP/E Cycle (times)SLC100,000UltraMLC30,000MLC3,0003D TLC (BiCS 3)3,000Table 2. P/E Cycle Times According to NAND TypePower Failure Protection Status: There are three power failure protection statuses.Voltage Stabilizer Trigger Count: The number of times the voltage stabilizer circuit hasAFhbeen triggered.Guaranteed Flush: This indicates the function is on/off status (0x01 Enable).Drive Status: Error code of power failure protection functions (0x00 Normal).CRC Error: Data error coding between the controller and host. If the CRC error count isDAhnot 0, the SATA signal is poor. Users should conduct a SATA signal trace from the host tothe device (PCB layout, SATA cable, and SATA connector).SSD Life Remaining, SSD Life Left: Presented as a percentage and calculated based onE7hthe average erase count and NAND reference erase count.𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝑬𝒓𝒂𝒔𝒆 𝑪𝒐𝒖𝒏𝒕𝐒𝐒𝐃 𝐋𝐢𝐟𝐞 𝐋𝐞𝐟𝐭 𝟏 ()%𝑵𝑨𝑵𝑫 𝒓𝒆𝒇𝒆𝒓𝒆𝒏𝒄𝒆 𝒆𝒓𝒂𝒔𝒆 𝒄𝒐𝒖𝒏𝒕EAhEBhTotal NAND Read: The total data size that the SSD controller reads from NAND flash.Sector unit size is 512 bytes.Total NAND Written: The total data size that the SSD controller writes to NAND flash.Sector unit size is 512 bytes.Table 3. SQFlash-Defined S.M.A.R.T. IDsSSD Lifespan PredictionAlthough the Endurance Check pie chart can be referenced to determinethe remaining product lifespan, the pie chart cannot predict the exactend-of-life (EOL). To help users predict the SSD lifespan both generally andunder specific operating conditions, Advantech’s SQFlash Utility softwarefeatures General Prediction and Endurance Estimation functions.General PredictionThe General Prediction function provides a general EOL prediction for theSSD. This prediction is based on the remaining SSD lifespan andcalculated using the following formula:5www.advantech.com

𝐄𝐬𝐭𝐢𝐦𝐚𝐭𝐞𝐝 𝐮𝐬𝐚𝐛𝐥𝐞 𝐭𝐢𝐦𝐞 (𝑭𝒍𝒂𝒔𝒉 𝑷/𝑬 𝑪𝒚𝒄𝒍𝒆 𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝑬𝒓𝒂𝒔𝒆 𝑪𝒚𝒄𝒍𝒆) 𝑺𝒂𝒇𝒆𝒕𝒚 �� 𝑬𝒓𝒂𝒔𝒆 𝑷𝒐𝒘𝒆𝒓 𝒐𝒏 𝒉𝒐𝒖𝒓𝒔The predicted EOL provided by the General Prediction function ispresented as a specific date and time, as shown in Figure 3. This date isdetermined based on the estimated usable time and the current date andtime. The Safety Factor acts as a buffer that allows users to prepare forSSD replacement. The Safety Factor value can be set as 0.8.Figure 3. SSD Estimated LifespanEndurance EstimationThe Endurance Estimation function provides an exact EOL prediction forthe SSD under specific operating scenarios. This function can assist userswith risk assessments and spare parts inventory management. Thepredicted EOL is determined by using the “Start” and “Stop” buttons tomeasure the average erase cycle generated by writing data within aspecified period of time. The Write Amplification and Terabyte Written(TBW) factors for the specific operating scenario can be calculated usingthe following formula:𝐒𝐒𝐃 𝐋𝐢𝐟𝐞𝐭𝐢𝐦𝐞 𝐄𝐬𝐭𝐢𝐦𝐚𝐭𝐢𝐨𝐧 𝐖𝐫𝐢𝐭𝐞 �� 𝐓𝐁𝐖 𝑭𝒍𝒂𝒔𝒉 𝑴𝒂𝒙𝒊𝒎𝒖𝒎 𝑷/𝑬 𝑪𝒚𝒄𝒍𝒆 𝑻𝒆𝒔𝒕 𝒕𝒊𝒎𝒆𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝑬𝒓𝒂𝒔𝒆 �� 𝑬𝒓𝒂𝒔𝒆 𝑪𝒚𝒄𝒍𝒆𝒔 𝑺𝑺𝑫 𝑪𝒂𝒑𝒂𝒄𝒊𝒕𝒚𝑯𝒐𝒔𝒕 𝑾𝒓𝒊𝒕𝒆𝑺𝑺𝑫 𝑪𝒂𝒑𝒂𝒄𝒊𝒕𝒚 𝑭𝒍𝒂𝒔𝒉 𝑷/𝑬 𝑪𝒚𝒄𝒍𝒆𝑾𝒓𝒊𝒕𝒆 ��*Average Erase Cycles: The average erase cycles added since the Start button was last pressed.*Host Write: The total amount of data written by the host to the SSD since the Start button was last pressed.When the “Start” button is pressed, data for the specific operating scenariois collected in real time, as shown in Figure 4. When the task or operationis completed, press the “Stop” button to obtain the results. A pop-upwindow displaying the text results will then appear, as shown in Figure 5.6www.advantech.com

Figure 4. Endurance Estimation Real-Time DataFigure 5. Test Results for a Specific Operating ScenarioRemote ManagementFor SSD status and lifespan monitoring, industrial application users mayhave one-to-many or many-to-one remote monitoring requirements. Inorder to meet diverse needs, Advantech’s WISE-PaaS/DeviceOn remotemanagement and monitoring platform was equipped with a Remote SSDMonitoring function. The WISE-PaaS/DeviceOn platform comprises twomajor components — a server-side system (the device manager) and adevice management agent. The server-side system communicates withdevices by using various protocols in order to provide both individual andbulk device control. Additionally, the server-side system facilitates remotemanagement of the software and applications deployed on devices, andcan be used to monitor/reset device applications or processes if necessary.7www.advantech.com

Figure 6. SQFlash Cloud-Based Management DiagramUsers can remotely monitor the SSD’s current operating status andremaining lifespan in real time from a PC or cell phone via theWISE-PaaS/DeviceOn web page. The WISE-PaaS/DeviceOn web pageprovides critical information, such as the Endurance Check results, ECCcount and good Block Rate, as shown in Figure 7.Figure 7. WISE-PaaS/DeviceOn Operation Interface8www.advantech.com

To allow users to understand the SSD status at a glance, theWISE-PaaS/DeviceOn platform also features a built-in dashboard. Userscan customize the dashboard layout to emphasize or highlight specific data,as shown in Figure 8.Figure 8. WISE-PaaS/DeviceOn SQFlash DashboardPredictive Maintenance QualityTo provide users with another means for quickly understanding the SSDstatus, the WISE-PaaS/DeviceOn platform is also equipped with aPredictive Maintenance Quality (PMQ) function. The PMQ value iscalculated based on the predicted EOL provided by the General Predictionfunction (see Section 2.2.1), with historical RMA and DOA failure modeanalysis used as the auxiliary criterion. The RMA and DOA failure modeanalysis procedures are regularly updated to optimize machine learningalgorithms. The PMQ user interface shows the current health status andlatest event notifications. For events that require special attention, the PMQinterface will display a warning message or issue a user notification, asshown in Figure 9.9www.advantech.com