Our conclusion based on our benchmark results is straightforward: The RAID array using different hard drives worked properly and offered performance that was comparable, but its performance was clearly behind that of a RAID array based on identical hard drives - all using the same system and RAID controller. The only real exception was the access time, which suffered substantially when different drives were used. Hence, only the array using homogenous drives would deliver maximum performance, which consists of a mix of quick access time, high read and write performance and good I/O benchmark results.
When comparing between identical RAID levels (RAID 0 or RAID 5), the throughput difference between a homogenous and a heterogeneous RAID array was anywhere between 10 and 25%. This is substantial if you need your RAID array to perform fast. For an average file server with only few people accessing data and with redundancy being the primary objective, the performance impact might be acceptable and good enough to get a business started.
However, knowing that 500 GB hard drive prices start at $90 today, we clearly recommend the clean solution, because you will not only get better performance, but you can be sure that the array should last long using all-new hard drives (especially if you use RAID and 24/7 optimized drives), while performance will be more consistent.
- RAID Scaling Charts, Part 1: 2-8 Drives Compared (RAID 0, 1)
- RAID Scaling Charts, Part 2: 2-8 Drives Compared (RAID 5, 6)
- RAID Scaling Charts, Part 3: 2-8 Drives Compared (4-128 kB Stripe Size)