About RAID

磁盘阵列（Redundant Arrays of independent Disks，RAID），有“独立磁盘构成的具有冗余能力的阵列”之意。原理是利用数组方式来作磁盘组，配合数据分散排列的设计，提升数据的安全性。磁盘 阵列是由很多价格较便宜的磁盘，组合成一个容量巨大的磁盘组，利用个别磁盘提供数据所产生加成效果提升整个磁盘系统效能。RAID5和RAID10在企业中使用最为广泛。

RAID Level Min Drives Protection Description Strengths Weaknesses Typical Apps 　RAID 0 2 None Data striping without redundancy Highest performance No data protection; One drive fails, all data is lost High End Workstations, Data Logging, Real-Time Rendering, Transitory Data 　 RAID 1 2 Single Drive Failure Disk mirroring Very high performance; Very high data protection; Very good on write performance High redundancy cost overhead; Because all data is duplicated, twice the storage capacity is required Operating Systems, Transactional Databases 　 RAID 5 3 Single Drive Failure Block-level data striping with distributed parity Best cost/performance for transaction-oriented networks; Very high performance, very high data protection; Supports multiple simultaneous reads and writes; Can also be optimized for large, sequential requests Write performance is slower than RAID 0 or RAID 1 Data Warehousing, Web, Archiving, Basic File Servers, Disk Backup 　 RAID 6 4 2 Drive Failure Same as RAID 5 with x2 Parity distributed across an extra drive Offers Solid Performance with the additional fault tolerance of allowing availability to data if 2 disks in a RAID group to fail. Recommended to use more drives in RAID group to make up for performance and disk utilization hits compared to RAID 5 Must use a minimum of 5 drives with 2 of them used for parity so disk utilization is not as high as RAID 3/5. Performance is slightly lower than RAID 5 High Availability Solutions, Mission Critical Apps, Servers with Large Capacity Requirements 　 RAID 10 4 1 Disk Per Mirrored Stripe (not same mirror) Combination of RAID 0 (data striping) and RAID 1 (mirroring) Highest performance, highest data protection (can tolerate multiple drive failures) High redundancy cost overhead; Because all data is duplicated, twice the storage capacity is required; Requires minimum of four drives Databases, Application Servers 　 RAID 50 6 1 Disk Per Mirrored Stripe Combination of RAID 0 (data striping) and RAID 5 (Single Parity Drive) Highest performance, highest data protection (can tolerate multiple drive failures) High redundancy cost overhead; Because all data is duplicated, twice the storage capacity is required; Requires minimum of four drives Databases, File Servers, Application Servers, 　 RAID 60 8 2 Disks Per Mirrored Stripe Combination of RAID 0 (data striping) and RAID 6 (Dual Parity Drives) Highest performance, highest data protection (can tolerate multiple drive failures) High redundancy cost overhead; Because all data is duplicated, twice the storage capacity is required; Requires minimum of four drives High Availability Solutions, Mission Critical Apps, Servers with Large Capacity Requirements

RAID Level	Number of Disks	Capacity	Storage Efficiency	Fault Tolerance	Availability	Random Read Perf	Random Write Perf	Sequential Read Perf	Sequential Write Perf	Cost
0	2,3,4,...	S*N	100%	none						$
1	2	S*N/2	50%							$$
2	many	varies, large	~ 70-80%							$$$$$
3	3,4,5,...	S*(N-1)	(N-1)/N							$$
4	3,4,5,...	S*(N-1)	(N-1)/N							$$
5	3,4,5,...	S*(N-1)	(N-1)/N							$$
6	4,5,6,...	S*(N-2)	(N-2)/N							$$$
7	varies	varies	varies							$$$$$
01/10	4,6,8,...	S*N/2	50%							$$$
03/30	6,8,9,10,...	S*N0*(N3-1)	(N3-1)/N3							$$$$
05/50	6,8,9,10,...	S*N0*(N5-1)	(N5-1)/N5							$$$$
15/51	6,8,10,...	S*((N/2)-1)	((N/2)-1)/N							$$$$$

Oracle and Raid

File Type	Preferred RAID Level	Comments
Control Files	RAID 1+0, RAID 0+1, RAID 0, No RAID	Control files are updated constantly so the quicker they can be accessed the better. Multiple control files should always be used whether you use RAID or not. When using no RAID or RAID 0 it is up to you to make sure that at least one copy of the control file is always available.
Online Redo Logs	RAID 1+0, RAID 0+1, RAID 0, No RAID	Similar access requirements to control files. Once again, the redo logs should always be multiplexed whether you use RAID or not.
Temporary Datafiles	No RAID, RAID 0	The main requirement here is rapid access, not reliability. If the datafile is lost it can simply be recreated as there is no data to restore.
Archived Redo Logs	RAID 1+0, RAID 0+1, RAID 0, No RAID	Always multiplex these, even when using RAID, if you can afford the space. If space is an issue, rely on RAID to provide redundancy. It's dangerous to use no RAID or RAID 0 without manually multiplexing them. If you lose archived redo logs you may have compromised your backup and recovery.
Rollback/Undo Datafiles	RAID 1+0 RAID 0+1	These files require constant I/O and must be protected. They cannot be mutliplexed by Oracle so let the hardware do it for you.
Datafiles	RAID 1+0 RAID 0+1 or RAID 5	Datafiles with heavy I/O requirements should use RAID 1+0 (or 0+1). It is the fastest and most secure option. If price is more of a consideration than performance, you can consider RAID 5. Most I/O operations to datafiles are buffered, with the physical writes happening in the background. As a result RAID 5 can be used to reduce costs without a significant impact on low performance systems.

Conclusion

• If you can afford it, use RAID 1+0 for all your Oracle files and you shouldn't run into too many problems. If you are using ASM, use this RAID level for all LUNs presented to ASM.
• To reduce costs move datafiles with lower access rates to RAID 5. If you are using ASM, this may mean defining multiple disk groups to hold different files.
• To reduce costs further move the other datafiles to RAID 5.
• To reduce costs further experiment with moving redo logs and rollback/undo tablespaces to RAID 5. Pick the combinations that work best for your system.
• If cost dictates it, move all your Oracle files on to RAID 5.

参考

Oracle and RAID

http://www.pcguide.com/ref/hdd/perf/raid/levels/comp-c.html

http://www.enhance-tech.com/press/raid-comparison-and-storage-systems.html