I have to agree with you - that article is
crap. As you say, he compares a 3-disk RAID 5 to a 4-disk RAID 10. Not only that, but:
- He notes the read performance of the RAID10 array as "4x" - however he does not note that this is dependent on the data layout algorithm, whether "near" or "far"; many implementations are "near", which will only involve two of the drives.
- The fault tolerance of RAID 10 is "1 disk per mirrored pair", not simply "1 disk".
- The RAID 5 array should have been 5 drives, not 3 if he assumed that the RAID 10 array was using the "far" data layout.@thread
Further more, out of the RAID toplogies in use today, none stand out as being better than any other. RAID 0
is the fastest, but has no fault tolerance. RAID 1
is the safest, but lacks performance. RAID 10
combines the best of the first two but incurs a minimum capital cost and energy penalty. RAID 5 & 6 have more flexible fault tolerance handling and can be grown organically, but can incur write penalties (which can be mitigated) and are probably more suited to "read mostly" applications, like media servers
Don't worry, I know I'm preaching to the converted. @OP
Out of all of the RAID implementations available in the Windows world, hardware RAID is the most reliable and best performing. The implementation on the cards is far superior to the "fake RAID" on the motherboard chipsets and the software implementation in the Windows OS.