SNIA-SDC23-Wells-Allison-Standardized-Storage-Telemetry.pdf

《SNIA-SDC23-Wells-Allison-Standardized-Storage-Telemetry.pdf》由會員分享，可在線閱讀，更多相關《SNIA-SDC23-Wells-Allison-Standardized-Storage-Telemetry.pdf（27頁珍藏版）》請在三個皮匠報告上搜索。

1、1|2023 SNIA.All Rights Reserved.Virtual ConferenceSeptember 28-29,2021Standardized Storage TelemetryEnabling Secure Fleet Monitoring and DebugSteven Wells Fellow Micron TechnologyMike Allison Senior Director Samsung Semiconductor2|2023 SNIA.All Rights Reserved.Current Solutions for Monitoring and De

2、bugSMART Multiple standardized SMART pages No Vendor Unique(VU)SMART.All remaining fields are reserved and shall be 0h OCP Datacenter NVMe Specification*does not allow VU log pages outside of specification(SEC-19*)without prior written consent by the customer.Telemetry Prior guidance(TEL-5*)was to u

3、se Data Area 1 for fleet monitoring but offered no payload guidance No guidance of Reason Identifier*payloadSecurity SEC-22*offers no guidance on what constitutes“Human Readable”*OCP Datacenter NVMe SSD Specification v2.0,July 2021Fleet monitoring requires many periodic IOs and no support for vendor

4、 unique detailsDebug is vendor uniqueVarying interpretations of Human Readable decode3|2023 SNIA.All Rights Reserved.New Standardized Telemetry StandardStandardizes the NVMe Telemetry Payloads in Data Area 1,Data Area 2 and Telemetry Reason Identifier Offers low latency/single command fleet monitori

5、ngData area 1Incorporates data from other useful SMART pagesAdditional standardized statistics supporting health monitoring and fleet balancing opportunitiesSupports vendor unique statistics and event FIFO capabilities in a standardized way Offers a standardized debug method post panicData area 2Rea

6、son Identifier(useful for segmentation)Offers a single,open-source OCP-NVMe-CLI decode generating human readable outputIncludes Telemetry String Log for VU statistics and eventsSingle command periodic fleet monitoring including VU detailsOpen industry standardized failure analysis with human readabl

7、e outputExample of standardized and VU extensible diagnostic and debug solution from automotive industry4|2023 SNIA.All Rights Reserved.Standardized Telemetry OverviewIssue DifficultyFleet Health MonitoringData Area 1Fleet Drive DebugData Area 2Reason IdentifierVU Debug logs+data area 3/4Physical Ac

8、cess to the driveStandardized Human Readable DecodeNot Decoded or VU Decode5|2023 SNIA.All Rights Reserved.Issue DifficultyFleet Health MonitoringData Area 1Fleet Drive DebugData Area 2Reason IdentifierVU Debug logs+data area 3/4Physical Access to the driveStandardized Human Readable DecodeNot Decod

9、ed or VU DecodeFleet Health MonitoringVU String Decode LogpageOCP-NVMe CLI Telemetry Plugin Human ReadableVendorDatacenter Security Inspection6|2023 SNIA.All Rights Reserved.Issue DifficultyFleet Health MonitoringData Area 1Fleet Drive DebugData Area 2Reason IdentifierVU Debug logs+data area 3/4Phys

10、ical Access to the driveStandardized Human Readable DecodeNot Decoded or VU DecodeStandardized Telemetry DebugVU String Decode LogpageOCP-NVMe-CLI Telemetry Plugin VendorDatacenter Security InspectionHuman Readable7|2023 SNIA.All Rights Reserved.Telemetry Layout SummaryData Area 1 for Health Monitor

11、ingNVMe and OCP SMART+Standardized and VU Telemetry Statistics+VU Event FIFO(s)Data Area 2 for DebugVU Event FIFO(s)Data Area 4Host InitiatedDevice InitiatedData Area 4 Unconstrained SizeVU Crash DumpsTelemetry HeaderStandardized Reason IdentifierLegend:Human ReadableLow IO LatencyHigh IO LatencyDat

12、a Area 332MB per NVMeEDTAS8|2023 Storage Developer Conference.Insert Company Name Here.All Rights Reserved.Specification Details9|2023 SNIA.All Rights Reserved.NVMe Telemetry Log PageData Area 4Data Area 3Data Area 2Data Area 1Header indicates the size of the Data AreasHeaderDefined by the NVM Expre

13、ss Base Specification 2.010|2023 SNIA.All Rights Reserved.OCP Telemetry Log Page FormatData Area 4Data Area 3Data Area 2Data Area 1HeaderFixed SizeStatisticsFIFOZero Filled PaddingFIFOOCP HeaderRelative to the start of the Data Area 111|2023 SNIA.All Rights Reserved.OCP Telemetry Log Page FormatData

14、 Area 4Data Area 3Data Area 2Data Area 1HeaderVersionStatistics InformationFIFO InformationNVMe SMART/Health Information log pageOCP SMART/Health Information Extended log pageProfile InformationStatisticsFIFOZero Filled PaddingFIFOOCP Header12|2023 SNIA.All Rights Reserved.OCP Telemetry Log Page For

15、matData Area 4Data Area 3Data Area 2Data Area 1HeaderEntry 0Entry 1Entry NStatisticsFIFOZero Filled PaddingFIFOOCP HeaderThe#of Entries isVendor Defined13|2023 SNIA.All Rights Reserved.OCP Telemetry Log Page FormatData Area 4Data Area 3Data Area 2Data Area 1HeaderStatisticsFIFOZero Filled PaddingFIF

16、OOCP HeaderDataIdentifierPersistenceNamespaceData SizeEntry 0Entry 1Entry NIdentifier can specify an OCP defined Statistic or a Vendor defined Statistics14|2023 SNIA.All Rights Reserved.OCP Telemetry Log Page FormatData Area 4Data Area 3Data Area 2Data Area 1HeaderStatisticsFIFOZero Filled PaddingFI

17、FOOCP HeaderIdentifierPersistenceNamespaceData SizeDataEntry 0Entry 1Entry NPersistence indicates the behavior of the data due to resets15|2023 SNIA.All Rights Reserved.OCP Telemetry Log Page FormatData Area 4Data Area 3Data Area 2Data Area 1HeaderStatisticsFIFOZero Filled PaddingFIFOOCP HeaderIdent

18、ifierPersistenceNamespaceData SizeDataEntry 0Entry 1Entry NNamespace indicates is the data is associated with a namespace and if so which one16|2023 SNIA.All Rights Reserved.OCP Telemetry Log Page FormatData Area 4Data Area 3Data Area 2Data Area 1HeaderStatisticsFIFOZero Filled PaddingFIFOOCP Header

19、IdentifierPersistenceNamespaceData SizeDataEntry 0Entry 1Entry NThe data size is dependent in the Identifier(i.e.,variable sized entries)17|2023 SNIA.All Rights Reserved.OCP Telemetry Log Page FormatData Area 4Data Area 3Data Area 2Data Area 1HeaderStatisticsFIFOZero Filled PaddingFIFOOCP HeaderEntr

20、y 0:OldestEntry 1Entry N:NewestThe#of Entries isVendor Defined18|2023 SNIA.All Rights Reserved.OCP Telemetry Log Page FormatData Area 4Data Area 3Data Area 2Data Area 1HeaderStatisticsFIFOZero Filled PaddingFIFOOCP HeaderDebug Type is a classification(e.g.,NVMe,PCIe,Resets,Vendor Specific)Debug Type

21、IdentifierDataVendor IdentifierVendor DataData SizeEntry 0:OldestEntry 1Entry N:Newest19|2023 SNIA.All Rights Reserved.OCP Telemetry Log Page FormatData Area 4Data Area 3Data Area 2Data Area 1HeaderDebug TypeIdentifierDataVendor IdentifierVendor DataData SizeEntry 0:OldestEntry 1Entry N:NewestIdenti

22、fier within the Class(i.e.,OCP Defined or Vendor Specific)StatisticsFIFOZero Filled PaddingFIFOOCP Header20|2023 SNIA.All Rights Reserved.OCP Telemetry Log Page FormatData Area 4Data Area 3Data Area 2Data Area 1HeaderDebug TypeIdentifierDataVendor IdentifierVendor DataData SizeEntry 0:OldestEntry 1E

23、ntry N:NewestData is the data Identifier within the Class(i.e.,OCP Defined or Vendor Specific)StatisticsFIFOZero Filled PaddingFIFOOCP Header21|2023 SNIA.All Rights Reserved.OCP Telemetry Log Page FormatData Area 4Data Area 3Data Area 2Data Area 1HeaderDebug TypeIdentifierDataVendor IdentifierVendor

24、 DataData SizeEntry 0:OldestEntry 1Entry N:NewestVendor Identifier and Vendor Data is optional additional vendor specific information for an OCP define Debug TypeStatisticsFIFOZero Filled PaddingFIFOOCP Header22|2023 SNIA.All Rights Reserved.Relative to the start of the Data Area 1Relative to the st

25、art of the Data Area 2OCP Telemetry Log Page FormatData Area 4Data Area 3Data Area 2Data Area 1HeaderStatisticsFIFOZero Filled PaddingFIFOOCP HeaderStatisticsFIFOZero Filled PaddingFIFO23|2023 SNIA.All Rights Reserved.Vendor defined definition ASCII stringGenerating Human Readable StringsData Area 4

26、Data Area 3Data Area 2Data Area 1HeaderStatisticsFIFOZero Filled PaddingFIFOStatisticsFIFOZero Filled PaddingFIFOOCP HeaderParse the Telemetry log page and for each Statistics or Event print:The dataThe ASCII string definitiono Hard code standards defined texto Lookup Vendor definedtext in the OCP S

27、trings log pageVendor Specific IdentifierOCP Strings log page24|2023 SNIA.All Rights Reserved.OCP Strings Log PageACSII text for the VU IdentifierRelative to the start of log pageIdentifierLengthOffsetRelative to the start of the ASCII StringsASCII StringsOCP Event Vendor Specific IdentifierString T

28、ableVendor Specific Event IdentifierString TableVendor Specific Statistics IdentifierString TableHeaderEntry 0:Smallest IDEntry 1Entry N:Largest ID25|2023 SNIA.All Rights Reserved.OCP Strings Log PageACSII text for the VU IdentifierRelative to the start of the ASCII StringsASCII StringsOCP Event Ven

29、dor Specific IdentifierString TableVendor Specific Event IdentifierString TableVendor Specific Statistics IdentifierString TableHeaderEntry 0:Smallest IDEntry 1Entry N:Largest IDIdentifierLengthDebugTypeOffset26|2023 SNIA.All Rights Reserved.OCP Strings Log PageACSII text for the VU IdentifierRelati

30、ve to the start of the ASCII StringsASCII StringsVendor Specific Event IdentifierString TableVendor Specific Statistics IdentifierString TableHeaderOCP Event Vendor Specific IdentifierString TableIdentifierLengthDebugTypeOffsetEntry 0:Smallest IDEntry N:Largest IDEntry 127|2023 SNIA.All Rights Reserved.Please take a moment to rate this session.Your feedback is important to us.