Strengths and weaknesses of Linux software RAID

I’ve been doing a lot of research on storage solutions for a couple of projects. I’ve been delving into RAID setups for large numbers of disks and trying to determine in what situations a hardware RAID controller is appropriate instead of Linux software RAID. I’ve also been considering various nested RAID levels and their value.

Linux Software RAID

Strengths:

• Highly configurable
• High performance
• Portable
• RAID 10 driver (in place of nesting that can use odd number of drives)
• Tools for analysis and reporting
• Free and Open Source

Weaknesses:

• Only RAID 1 is bootable (technically a problem of the bootloaders)
• Only RAID 4,5,6 are growable (and RAID 1 but that’s of limited use)
• Cannot reshape arrays (change RAID level)

Those weaknesses can be overcome by a RAID controller but possibly by sacrificing some of the strengths. Still both the issue that arrays cannot be expanded nor booted might make a hardware controller more appealing, despite the cost and other trade-offs.

RAID designs for large numbers of disks

• RAID 5 - probably the most popular design as it is relatively simple and can survive 1 drive failure, capacity is N-1, has strong read performance but not write performance, performance is poor when degraded or rebuilding.
• RAID 6 - like RAID 5 but can survive 2 drive failures, capacity is N-2, performance is slightly worse than RAID5, performance is poor when degraded or rebuilding.
• RAID 1+0 (“stripe of mirrors”) - improves upon performance over RAID 5/6 at the cost of capacity which is N/2, it can withstand at least 1 drive failure and potentially more (up to N/2).
• RAID 0+1 (“mirror of stripes) - should never be used as it offers nothing over RAID 1+0.
• RAID 0+5 and RAID 5+0 - both of these attempt to improve performance over RAID 5/6 without losing as much capacity as RAID 1+0. These are not common but should be considered with large numbers of drives is RAID 5 is too slow and RAID 10 is too small.