8 summary of raid performance characteristics – IBM SG24-4576-00 User Manual
Page 45

1.6.6.8 Summary of RAID Performance Characteristics
RAID-0:
Block Interleave Data Striping without parity
•
Fastest data-rate performance
•
Allows seek and drive latency to be performed in parallel
•
Significantly outperforms single large disk
RAID-1:
Disk Mirroring/Disk Duplexing and Data Strip mirroring (RAID-1,
Enhanced)
•
Fast and reliable, but requires 100% disk space overhead
•
Data copied to each set of drives
•
No performance degradation with a single disk failure
•
RAID-1 enhanced provides mirroring with an odd number of drives
RAID-2:
Bit Interleave Data Striping with Hamming Code
•
Very fast for sequential applications, such as graphics modelling
•
Almost never used with PC-based systems
RAID-3:
Bit Interleave Data Striping with Parity
•
Access to all drives to retrieve one record
•
Best for large sequential reads
•
Very poor for random transactions
•
Poor for any write operations
•
Faster than a single drive, but much slower than RAID-0 or RAID-1 in random
environments
RAID-4:
Block Interleave Data Striping with one Parity Disk
•
Best for large sequential I/O
•
Very poor write performance
•
Faster than a single drive, but usually much slower than RAID-0 or RAID-1
RAID-5:
Block Interleave Data Striping with Skewed Parity
•
Best for random transactions
•
Poor for large sequential reads if request is larger than block size
•
Better write performance than RAID-3 and RAID-4
•
Block size is key to performance, must be larger than typical request size
•
Performance degrades in recovery mode (when a single drive has failed)
Table 7. Summary of RAID Performance Characteristics
RAID Level
Capacity
Large Transfers
High I/O Rate
Data Availability
Single Disk
Fixed (100%)
Good
Good
1
RAID-0
Excellent
Very Good
Very Good
Poor
2
RAID-1
M o d e r a t e ( 5 0 % )
Good
Good
Good
RAID-2
Very Good
Good
Poor
Good
RAID-3
Very Good
Very Good
Poor
Good
RAID-4
Very Good
Very Good
Poor
Good
RAID-5
Very Good
Very Good
Good
Good
Note:
1
The MTBF (mean time before failure) for single disks can range from 10,000 to 1,000,000 hours.
2
Availability = MTBF of one disk divided by the number of disks in the array.
30
NetWare Integration Guide