1 INTRODUCTION
2 BACKGROUND
2.1 Disk Terminology
2.2 Data Paths
2.3 Technology Trends
3 DISK ARRAY BASICS
3.1 Data Striping and Redundancy
3.2 Basic RAID Organizations
3.2.1 Non-Redundant (RAID Level 0)
3.2.2 Mirrored (RAID Level 1)
3.2.3 Memory-Style ECC (RAID Level 2)
3.2.4 Bit-Interleaved Parity (RAID Level 3)
3.2.5 Block-Interleaved Parity (RAID Level 4)
3.2.6 Block-Interleaved Distributed-Parity (RAID Level 5)
3.2.7 P+Q Redundancy (RAID Level 6)
3.3 Performance and Cost Comparisons
3.3.1 Ground Rules and Observations
3.3.2 Comparisons
3.4 Reliability
3.4.1 Basic Reliability
3.4.2 System Crashes and Parity Inconsistency
3.4.3 Uncorrectable Bit-Errors
3.4.4 Correlated Disk Failures
3.4.5 Reliability Revisited
3.4.6 Summary and Conclusions
3.5 Implementation Considerations
3.5.1 Avoiding Stale Data
3.5.2 Regenerating Parity after a System Crash
3.5.3 Operating with a Failed Disk
3.5.4 Orthogonal RAID
4 ADVANCED TOPICS
4.1 Improving Small Write Performance for RAID Level 5
4.1.1 Buffering and Caching
4.1.2 Floating Parity
4.1.3 Parity Logging
4.2 Declustered Parity
4.3 Exploiting On-Line Spare Disks
4.4 Data Striping in Disk Arrays
4.5 Performance and Reliability Modeling
5 CASE STUDIES
5.1 Thinking Machines Corporation ScaleArray
5.2 StorageTek Iceberg 9200 Disk Array Subsystem
5.3 NCR 6298
5.4 TickerTAIP/DataMesh
5.5 The RAID-II Storage Server
5.6 IBM Hagar Disk Array Controller
6 OPPORTUNITIES FOR FUTURE RESEARCH
6.1 Experience with Disk Arrays
6.2 Interaction among New Organizations
6.3 Scalability, Massively Parallel Computers, and Small Disks
6.4 Latency
7 CONCLUSIONS
8 ACKNOWLEDGEMENTS
9 ANNOTATED BIBLIOGRAPHY