Which RAID type is better?

 

Which one is better? RAID 5 + 1 Hot-spare or RAID 6 for your storage configuration:

It’s a very common question: Which types of array configuration we must choose, whenever we want to configure storage devices like SAN storage or even local disks of an enterprise server. To answer this simple but important question, we need to dive into the details of three vital factors: Storage capacity, the overall performance (IOPS), and, data protection mechanism (fault tolerance). First of all, we need to investigate the characteristics of the most popular RAID types: 0+1, 1+0, 5, and 6.

In cases of using mirror RAIDs (1+0 or 0+1) despite increasing data-protection rate and good performance, we lose 50% of storage capacity. So, in medium-scale environments, it’s a massive limitation and can be a bottleneck on storage design. In most design documents with parity-bit array configuration like RAID 5 (single parity tier, and single disk fault tolerance) and RAID 6 (two parity tiers, and simultaneous two disks fault tolerance), there is a good recommendation: Configure one or multiple disks of the storage device as the Hot Spares of your arrays. So, if a failure happens one of the spare disks automatically replaced with the failed one. (After replacement of healthy disk, it will be select as the next hot-spare) While this is not a bad idea but what will happen if multiple concurrent disk failures or second failure occurs before the parity-bit structure rebuild the array? In this situation, unfortunately, we will lose everything in the failed array, So, now what can we do?! The answer is nothing, just re-design the storage infrastructure!

By considering the same circumstances for the storage space in a comparison between RAID6 and RAID5+1spare (In both of them, storage of two disks is sacrificed for their array construction) we should choose one of them for our storage array, with respect to the protection metric or the performance. But after executing many performance tests (For example, computing latency percent of the read/write operations) no significant difference was observed. Although I don’t want to say there is no overhead for calculation of the second parity-tier in RAID6. However, in front of the protection factor, additional mathematical operation of the RAID6 does not degrade the performance rate effectively. So, if we need to choose between RAID5+1spare and RAID6, and select the best options for our storage infrastructure, I say you should specify first of all “which one is more important? overall IOPS (performance) or fault tolerance (data protection)”

If you choose the first option, I will say configure RAID5 for all arrays, with one or multiple hot-spare disks based on needs and your design, while if you want the second one, I will suggest RAID6 is the best choice you ever made. Because traditionally we made most of the RAID structures to prevent data loss, finally I can conclude RAID6 is the winner of this game because the storage volumes can tolerate double disk failures. Especially in the case of using large capacity disks (more than 1TB space) risk of second disk failure before finishing the rebuild operation is very high and It’s an important point you should always consider.