1. RAID: High-Performance, Reliable Secondary Storage Mei Qing & Chaoxia Liao Nov. 20, 2003

2. Outline  PART I: Basic RAID  Introduction  RAID organization  PART II: Performance Improvement  Performance Analyze  RAID 5 Improvement  Conclusion  Limitation  Future Work

3. PART I Basic RAID

4. Introduction  Background  What is RAID  RAID stands for Redundant Array of Inexpensive Disks  Driving force of RAID  Faster Microprocessors  Dramatic increase in the amount of data needed to serve and store  Storage technologies is very expensive

5. Intruduction  Advantage of RAID  Redundancy  Mirror  Parity distribution  Increased Performance  Striping  Lower Costs

6. Basic RAID Organization  RAID levels 0 to 5  RAID 0 (Non-redundant)  RAID 1 (Mirrored)  RAID 2 (Memory-Style ECC)  RAID 3 (Bit-Interleaved Parity)  RAID 4 (Block-Interleaved Parity)  RAID 5 (Block-Interleaved Distributed-parity)

7. RAID Level 0  Advantage & Disadvantage  Good performance on read  Low cost  No redundancy  Simple Design  Easy to implement  Recommended Applications  Video Production and Editing  Image Editing  Any application requiring high bandwidth

8. RAID Level 1  Advantage & Disadvantage  Fast on Read  Slow on Write  Very good redundancy  High Cost  Recommended Applications  Accounting  Payroll  Financial  Any application requiring very high availability

9. RAID Level 2 Hamming Error Correction Code

10. RAID Level 3  Advantage and Disadvantage  High efficiency  Good read and write  Data access in parallel  Recommended Applications  Video Production and live streaming  Image Editing  Video Editing  Prepress Applications  Any application requiring high throughput

11. RAID Level 4

12. RAID Level 5  Advantage & Disadvantage  Best small read, large read and large write performance  Overcome the bottleneck of the parity  Good reliability  Inefficiency small write because of the overhead of distributed parity  Recommended Applications  File and Application servers  Database servers  WWW, E-mail, and News servers  Intranet servers  Most versatile RAID level

13. Further Develop on RAID  Multiple Nested RAID  RAID level 1+0  RAID level 0+1  RAID level 53  RAID level 0+5

14. RAID Level 1+0  Advantage & Disadvantage  Good fault tolerance  Good rebuilt performance  Good read and write performance  Not require parity calculation  Recommended Applications  Database server requiring high performance and fault tolerance

15. RAID Level 53

16. RAID Implement  Hardware RAID  More efficiency  Operating system independent  Highly fault tolerance  Incompatible  Expensive  Software RAID  Run on the server’s CPU  Directly depends on server’s CPU speed  Occupy host system memory and CPU operation, degrading CPU performance  Cheap

17. PART II Performance Analyze and Improvement

18. Performance Analyze  Aspects:  Performance and cost  Metrics to evaluate performance  Implementation and Configuration  Reliability

19. Reliability Analyze  Reliability  Fault Tolerance  Availability

20. Tradeoff  Performance, Reliability and fault tolerance concern

21. Performance Summary RAID 0RAID 1RAID 5RADI 10 ReadHigh2XHigh WriteHigh1XMediumHigh Fault tolerance NoYes Disk utilization HighLowHighLow Key problems Data lost when any disk fails Use double the disk space Lower throughput with disk failure Very expensive, not scalable Key advantages High I/O performance Very high I/O performance A good overall balance High reliability with good performance

22. Implementation Consideration  Avoiding Stale Data  When disk fails  When invalid logic sector reconstructed  Regenerating Parity after system Crash  Inconsistent mark  Parity regenerated  Operating with a failed disk

23. RAID Level 5 Analyze  Importance  Have the best small read, large read,and large write performance.  Allowing somewhat better parallelism in a multiple-transaction environment  Good fault tolerance  Widely Used

24. RAID 5 Performance Improvement  Improvement on small write  Buffering and caching  Write buffering  Read caching  Floating Parity  Parity logging

25. Advanced topic  Advanced techniques used in the design of redundant disk arrays.  Declustered Parity  Exploiting On_line spare disks  Data stripping disk arrays  Performance and reliability modeling

26. Conclusion  Advantage  Higher Data Security  Fault Tolerance  Improved Availability  Increased, Integrated Capacity  Improved Performance  Disadvantage  High cost for business

27. Limitation  How to evaluate whether you should use RAID  Not referring some techniques for back up  Risk to use RAID

28. Future Work  Be careful to use RAID  Decide which level of RAID need to be used.  Make a decision on implementation of RAID, Hardware RAID or Software RAID  Lower the cost

29. Questions