Before we begin it must be understood that the RAID number does not determine the quality of the RAID level. The RAID number just represents a type of service and each option is valid for each type of server and environment. Based on your knowledge of your infrastructure, you should be able to determine which RAID is the most cost/performance effective for you. Of course that’s what this article will also help with, understanding the pros and cons of the different RAID levels. So, now that we’ve made this clear, let’s start explaining the different ‘Standard’ RAID levels.
RAID 0 – This is the step brother of the RAID, also called ‘stripping’. If you remember, we discussed that RAID stands for ‘Redundant Array of Independent Disks’. RAID 0 is a combination of all the drives chosen without any fault tolerance; not exactly part of the definition of the word redundant as there isn’t any. The reason it is categorized is because it does cluster several drives together into one logical drive. Since there is not fault tolerance, the total capacity of the logical drivers is a sum of all the drives together.
Since the data is stripped across multiple drives simultaneously, the performance is improved dramatically and this RAID is chosen when performance is more important than data integrity.
The formula is: Total capacity = number of drives * capacity of each drive.
RAID 1 – This is a very popular RAID level called ‘mirrored’ and it does just that. It mirrors the data of one drive onto the other drive. This RAID is limited to only 2 drives and takes half of the capacity for duplication. The performance is slower as the controller needs to write the data twice but data integrity is kept in case of a failure.
The formula is: Total capacity = total capacity / 2
RAID 5 – Now this is where it starts becoming tricky. RAID 5 is a block-level striping with distributed parity. Huh? Exactly! It consists of a minimum of 3 drives and basically provides fault tolerance by creating one drive for parity. This creates a distributed XOR bit. Upon distribution of the data among the drives, the data is divided between the drives and the controller creates a parity for the XOR which is a formula that combines the data that was distributed prior to the bit into a special character and saves that character in its place. Now remember, this is distributed so it will be all over the 3 drives (not putting all the eggs in one basket, as it were). The XOR formula could be a nugget of its own so just understand that the parity drive that provides the fault tolerance consists of a specialized character that, upon a drive failure, can be used to understand which piece of the data was lost with the failure. It provides a fault tolerance of one drive. The performance is not the best because the focus is on data integrity and stability.
The formula is: Total capacity = (number of drives – 1) * capacity of one drive
RAID 6 – This is the same as RAID 5 just provides double parity and can tolerate 2 failures. If RAID 5 had an XOR parity table, RAID 6 provides that and introduces the Galois field parity which is way more complicated. This provides a fault tolerance of 2 disk failures, therefore better fault tolerance.
This requires a minimum 4 drives and will take 2 drives for both of the parities. This is a perfect RAID for storage units where extra fault tolerance is required and the investment in additional parity drives is necessary.
The formula is: Total capacity = (number of drives – 2) * capacity of one drive
It is important to understand that there are more RAID level, such as RAID 2, 3, and 4, but those all can be outperformed by RAID 5 therefore they are usually not supported in servers. It is very uncommon to encounter one of those types and the ones I mentioned are the ones you will be hearing about or needing to know of.
As I said, the hybrid levels will be discussed in a future nugget.
Thank you for reading and feel free to leave me questions and I will answer them as quickly as I can.
No comments:
Post a Comment