驱动器再生技术的实际应用：增强数据中心的容错能力 -.pdf

1、Seagate TechnologyCurtis Stevens StrategistDave Craton Principal Product ManagerDrive Regeneration in Action:Enhancing Fault Tolerance in DatacentersStorage1.Datacenter storage challenges with fault tolerance2.Expanding HDD fault tolerance from sector to head/surface level3.Data-Safe Drive Regenerat

2、ion/Depop OptionsAgendaKey Datacenter Challenges ReplacingReplacingor rebuilding drives is or rebuilding drives is costlycostly technicians,shipping,processing,network impact,availability downtime etc.Millions of drives are shredded each Millions of drives are shredded each year year keeping drives

3、in use longer improves operational carbon impact,more sustainable than shredding or recycling.Standard Enterprise HDD Reliability:2.5M MTBF=0.35%Annualized Failure Rate1.75%CFR(5-year warranty)8760 hours/yearEnterprise HDD Reliability-BaselineFailure RateTimeEarly Life FailuresLow Constant Fail Rate

4、Wear-out FailuresDrives in a server or storage system in the datacenter may be rejected at a higher rate than baseline.Examples:System-wide offline events,performance variations,excursions,etc.Some heavy datacenter workloads and high temperatures may accelerate some HDD fail modes.Typical rejection

5、target 1%/yrCategorizing Datacenter HDD RejectionsGROUP 1:NTFNo trouble found with a return to service recommendation.Some portion of these may fail again and land in a category below.GROUP 2:SINGLE-HEAD FAILURESMay be resolved by Drive Regeneration/Depop.Can represent 40%+of failures.GROUP 3:DECOMM

6、ISSIONTypically the smallest group of failures recommended for decommissioning.On 20 head/10 disk drives,95%+of the good capacity may be preserved rather than discarded.Hundreds or thousands of rebuilds may be significantly reduced or prevented entirely.Methods for Managing Hard Drive FailuresHDD Te