Server Talk Episode 3: An Intro to RAID

Share on Google+Tweet about this on TwitterShare on FacebookShare on LinkedInPin on PinterestEmail this to someone

Welcome to our third installment of “Server Talk.” In this podcast, Alexey and Mike explore the topic of RAID: what is it and why is it important for your applications or business? Make sure to leave comments, Like, and Share us.

(Download Transcript)

Topics Covered:

  • ICC News: New website!
  • Industry News: SSD Shortage.
  • What does RAID do?
  • Breakdown of popular RAID levels (0, 1, 5, 10, etc.)
  • The balance of performance / redundancy / efficiency in RAID
  • How to choose a RAID configuration
  • The importance of RAID controllers

The term “RAID” stands for “redundant array of independent disks.” It refers to storage configurations which combine multiple disk drive components into effectively a single unit and then divide and replicate data among these multiple drives. The purpose of this to increase data redundancy and performance. RAID is an example of storage virtualization, as the entire array can be accessed by the operating system as one single drive (even though there are multiple physical drives).

Different configurations are named by the word RAID followed by a number. Below is a breakdown of the more popular RAID schemes, with visual diagrams and excerpts from the podcast.

RAID 0

Alexey: Alright, so RAID 0 – what that basically means is taking two hard drives and “striping” them. What “striping” them means is just taking the data and writing across both of them (the drives). And the performance – it increases a lot but also, your redundancy… it’s completely gone. What that means is that if you lose a disk, you’re screwed.

RAID 0 Diagram

Mike: Okay, so is this a common configuration?

Alexey: Not at all unless you have backups and just straight up looking for performance.

RAID 1

Alexey: The next configuration is a RAID 1, just called a “mirror”. Basically you have two hard drives and they’re an identical copy of each other. Performance is not as good as the RAID 0 but redundancy is much better because you have a full copy.

RAID 1 Diagram

RAID 5

Alexey: RAID 5 is two drives plus a parity, which means it’s a minimum of three drives. And what a parity is, it’s just the way that the RAID calculates the information that it runs. In simple words of explaining it… this is really dumbed down but it’s made just for simple purposes… imagine you’re writing information in three different areas. You knew what came before; you knew what came after the drive you’re missing. So imagine you’re writing the information in three different areas and imagine you lose one of those writes, say the one in the middle. Well… you knew the information that come before; you knew the information that came after; you can recover that missing information.

Mike: Okay, so the system just has some sort of internal logic to figure out missing pieces of the connective tissue between the two other drives?

Alexey: That’s correct. It’s actually a math calculation just about where it writes that parity.

RAID 5 Diagram

Alexey: It gives you the redundancy so you can lose a disk and it gives you more performance than a mirror. What RAID 5 also allows you to do is grow it aside from just two disks. Whereas a mirror is only two disks, RAID 5 you can grow up to as many disks that the RAID controller can handle. Usually, you don’t wanna go… usually there will be like thirty disks or five to thirty disks and if you lose one of them you can recover.

RAID 6

Mike: It’s an upgrade from RAID 5?

Alexey: Yup… it’s called a double parity. So, basically same concept but now you have two parities, you can lose two disks. And it’s a minimum of four disks in that RAID technology… it’s a little bit slower but a little bit more redundancy.

RAID 6 Diagram

RAID 10

Alexey: The last, most popular RAID technology is called RAID 10. RAID 10 consists of RAID 1 and RAID 0. Basically, you have two RAID 1s and they’re striped together with a RAID 0 giving you the redundancy of a RAID 1 and the performance of a RAID 0.

Mike: Okay, I’m just gonna have to imagine that it works because I understand each one individually but when… the idea of putting them together I conceptually don’t really understand. Can you go just a little bit more or am I…? It’s something that I need to study more.

Alexey: Sure. Imagine… Let’s draw a few different blocks. Let’s draw four different blocks with two on the left. I want to be blue color blocks.

Mike: Okay.

Alexey: The two on the right are gonna be yellow color blocks.

Mike: Okay.

RAID 10 Quick Mockup

Alexey: The different blocks are different mirror sets. So those are the drives. So if you lose one of the blue blocks, you’re still up there with another blue block. If you lose one yellow block, you’re still there with the other yellow block.

Mike: Got you.

Alexey: So now, we have these two sets of blocks. And now, we’re gonna stripe them together like a RAID 0 to give us that performance. Remember RAID 0 gives us that performance?

Mike: Yeah.

Alexey: So you have the redundancy of a mirror for each block set and then you have the performance with the stripe across the different blocks

RAID 10 Diagram

Listen to the podcast to learn more about how to choose a RAID configuration, the performance / redundancy / efficiency balance between the different RAID levels, and about the importance of RAID controllers. Also, be sure to check out our popular RAID calculator application (which also provides details on the costs and benefits involved with each RAID level).

Thanks for listening! Be sure to leave comments, Like, and Share us.

Comments

  1. Keep in mind that bit errors can mess with the RAID reliability.
    In the event of a drive failure, the rest of the drives have to be read 100% from start to finish. Any read error during the rebuild will cause complete data loss.

    http://www.techrepublic.com/blog/the-enterprise-cloud/how-to-protect-yourself-from-raid-related-unrecoverable-read-errors-ures/

    Depending on your drive’s specified unrecoverable read error rate and the total storage capacity, the probability of failed rebuild is worth paying attention to.

    http://www.lucidti.com/zfs-checksums-add-reliability-to-nas-storage

Leave a Reply