Right then, one of the housekeeping staff – a busty lady – handed him a note.
Had he scored? With a twinkle in his eye, he opened the note. On it was scrawled:
“To understand RAID, you’ll need to look at stability, speed and capacity.”
Stability tells us how well the RAID will remain active, and how well it can heal (rebuild data) itself once a drive fails.
Speed tells us how fast the RAID array will read and write data.
Capacity tells us how much of the total capacity needs to be sacrificed to fulfil a particular RAID architecture. A 1 GB drive in RAID does not always give 1 GB.
“Boys,” the detective announced, “we’ve hit the mother lode.”
RAID 1 needs at least two drives. If drive one holds X amount of data, drive two will have the same data copied to it – a duplication.
This duplication is called Mirroring.
If one drive fails, the other has all the information, and the system continues functioning.
RAID 1 is the simplest RAID architecture.
What’s the difference between two drives in RAID 1 and two drives having the same data but are JBODs?
By assigning the RAID 1 architecture to the two drives, you’re telling your computer that these two drives will behave like one drive.
When one drive fails, the RAID Controller will ensure the data is read from the second drive and all is normal. The Computer will never know what took place inside the RAID array. All it cares about is its data getting read or written.
In JBODs, on the other hand, when one drive fails, you’ll physically have to notify the computer or software to switch over to the other drive, and probably start from scratch as well.
As you may have noticed, RAID 1 does not use parity.
The advantages of RAID 1 are that it is simple to implement, and is compatible with many different kinds of systems. It gets along with almost everybody. Also, it can theoretically double the read rate of each drive. I say, theoretically, because it requires a special bit of juggling by the Controller to make this happen, and not all do.
The disadvantages of RAID 1 are that it can only write as fast as each individual drive’s write rate. Also, due to the issue with one to one mirroring we discussed in the last chapter, we’ll need at least double the capacity for each block of data. Adding more drives just makes it worse. Each drive will have the same information as the first drive. Five drives is five copies, with a space efficiency of 1/5 or 20%.
Oops, that’s why we need other levels of RAID.
RAID 1 is great when capacity isn’t a concern, two drives are good enough and the drive isn’t used much for writing, only reading. Within these limitations, it is king.
If you need only two drives, and are only reading source footage for your NLE or workstation, then nothing beats a RAID 1 with the right controller.
In the last chapter we also mentioned a RAID that is not really a RAID. That’s RAID 0.
RAID 0 is a wily character, and wasn’t easy to find. When the detectives entered the domain of RAID 0, they encountered a whole new level of weird.
From afar RAID 0 looks like RAID 1. Two drives, same nunchucks, no big deal. Nope. They are as different as male and female.
In RAID 0, data is written to both drives simultaneously, filling both of them up. They don’t have the same information.
RAID 0 has twice the capacity of RAID 1.
“You’re no good,” the detectives thought about RAID 0. “What good is a RAID that doesn’t have redundancy? You’re a blotch to your name.”
“Don’t judge me so quickly, detective,” RAID 0 answered. “I have a skill nobody else has. I am twice as fast.”
“That’s BS. Data is written to one drive, then another. How can you be twice as fast?”
The data written to RAID 0 is split into half. We have seen how the file system splits files into blocks.
The first block is written to drive one, and the second to drive two. The third is written to drive one, and the fourth to drive two, and so on, until all the data is written.
This way, no drive has to wait around, and as long as the computer can push data through the system at twice the speed the RAID 0 array can read and write at that speed.
“What about backup?” asked a sharp detective.
“I don’t have backup. I am built for speed.”
The great advantage of the RAID 0 array is that it can theoretically increase the total speed of the array depending on how many drives it has. If the array has two drives, the speed is doubled. If the array has ten drives, the speed is ten times the speed of a single drive!
RAID 0 leverages the speed limit imposed on data transfers due to hard drives and turns it on its head.
One tortoise can only nibble at a whale’s ears. Many tortoises strung together can eat the whale whole. With RAID 0, you can theoretically build a drive array that can out pace the entire computer or network system, so that your hard drives are no longer a liability.
The cutting up of data into pairs for writing or reading, as in the case of RAID 0, is called Striping.
All RAIDs do at least one of three things – Mirroring, Striping and Parity. This is what sets them apart from JBODs.
RAID 1 has no striping or parity. RAID 0 has no mirroring or parity.
The biggest disadvantage of RAID 0 is that if one drive fails, all is lost. Since there’s no redundancy at all, you lose all your data in that array, and will need to start from scratch.
Before you complain, remember that this is very similar to a single hard drive failure. However, one would be foolish to expect redundancy from such a system. RAID 0 is used for the best read and write speed, nothing else.
Which makes it the best candidate for real-time uncompressed media streaming – as long as everything else can keep pace.
It didn’t take long for people to realize if RAID 0 and RAID 1 were the gods of the RAID world, surely they can get along together.
What if we could have both mirroring and striping – speed and redundancy? How do we do it – RAID 0+1 or 1+0? Let’s look at that next.
Links for Further Research: