Shulou(Shulou.com)06/01 Report--

How to interpret RAID 2.0, many novices are not very clear about this, in order to help you solve this problem, the following editor will explain for you in detail, people with this need can come to learn, I hope you can gain something.

Today, I would also like to talk about a force of high-end storage, RAID 2.0, which has recently been boiled over by Huawei HVS, which is magical and puzzling.

Let me start with a quack of high-end storage.

It is said that a long time ago (don't throw rotten eggs, storytelling is like this. ), King L has a beautiful Princess D (representing database DB), especially likes to eat watermelon fruit plate (on behalf of LUN seen by the host), and eats an amazing amount of food, which can eat 2 at a time. Princess D has a lazy habit of eating fruit. Just like the Thais, the fruit is cut into small pieces (Extend) and then spelled into a LUN and eaten with toothpicks. Princess D also has a small temper. If a small piece of the fruit plate is bad, the whole fruit plate will be thrown away (representing data loss). King L spoiled Princess D so much that he hired EMC/IBM/HDS to do a long-term job, specializing in making fruit plates. There is a distance between the melon land and the palace. They pick three watermelons every day, two for eating and one for backup. In case the melon is broken, it will be made up (for RAID 5). In this way, they work hard in the palace and melon land every day, and when they encounter bad melons, they still have to go back to the melon fields to prepare melons (reconstruction), which is very hard. Princess D grew up day by day, eating more and more, and melons getting bigger and bigger. But the problem is that every time they encounter a bad melon, it takes them 10 hours to move the melon, because the watermelon is too heavy to walk fast on the road. Princess D couldn't stand it any longer and asked King L to remove them and replace them with 3PAR and HW,3PAR and HW. They thought, why do I go to the palace and then cut melons? I can cut melons into 20 large pieces (Chunk), find 20 people to move to the palace and then cut them into small pieces (Extend) and then put them into LUN. If you find that the melon is broken when you get to the palace, can't you just send someone back to get a small piece to replace it? In this way, the only melons that need to be moved are the original 1apace 20. The melons are light and can run on the road, so even if the melons are broken, they will be changed back in half an hour. Princess D was very satisfied and lived in peace until one day HW fell ill after overtime, and 3PAR happened to fall in love with HP and asked for leave, so he found a young man called XIV to do a temporary supervisor. This guy was more wasteful. Every time he picked four melons, cut them all into small pieces (Extend) in the melon land, and found a group of children to send directly to the palace. Of course, it is good for Princess D, but only half of it can be eaten at a time, and the other half is thrown out, which is a pity. Later, because of her beauty, XIV was raised by IBM, which is later said.

All right, that's the end of the story. Do you know what's going on with RAID 2.0? Also throw rotten eggs, no, let me tell you a good story, I was just telling a story.

As you know, the traditional RAID 5 process goes like this:

Select a few hard drives-"make RAID 5"-"create LUN- based on capacity" to map to the host (for convenience, let's call this RAID 1.0)

It is said that EMC/IBM/HDS, the mainstream high-end manufacturer at that time, was all this plan. In this way, if a disk is broken, only the hard disk of this disk group can participate in the reconstruction. At that time, the hard drives were generally dozens of gigabytes, and all of them were FC disks, so the problem was not serious.

But now the high end has introduced the SATA disk, now the watermelon, no, is the hard disk is getting bigger and bigger, therefore, when a hard disk is broken, only these hard drives participate in the reconstruction, the reconstruction time 1TB takes 10 hours, if it is the SATA disk of 4TB, it is even more unimaginable.

In addition to refactoring time, there is a big problem with RAID 1.0, which is performance. A LUN can only be read and written on one disk group, so that new media such as SSD added later will not work. But EMC/IBM/HDS has been accumulated in RAID 1.0 for more than a decade, and RAID is the foundation of all software, so they don't dare to rewrite code easily. What shall I do?

Yes, you can combine multiple RAID groups into a pool and cut it again (striped):

Select several hard drives-"make RAID 5"-select multiple RAID 5 to form a pool-"cut into equal small chunks of Extend-" and select Extend to form LUN, which is mapped to the host (for convenience, let's call this RAID 1.5)

RAID 1.5 solves the performance problem very well, because a LUN reads and writes across many hard disks at the same time, and the LUN can contain multiple RAID groups, that is, there can be multiple disk media, and automatic tiered storage can be achieved. However, because the RAID group is still based on the hard drive, the hard drive is broken, and only a few hard drives from one RAID group participate in the reconstruction, so the reconstruction speed is still the same as that of RAID 1.0.

3PAR and Huawei have little historical baggage, so they use block virtualization technology RAID 2.0 to solve this problem (3PAR is internally called FAST RAID). The idea of RAID 2.0is to cut a few knives before doing RAID, put the watermelon (stop fighting, don't play too many games), oh, is the hard disk cut into a lot of equal chunks (Chuck), then take Chunk as a unit to do RAID 5 (form CKG), and then cut CKG into smaller pieces (Extent), randomly or according to certain rules to extract a lot of Extend to form LUN, mapping to the host.

Select all hard drives-"all cut into large chunks of Chuck-" to make RAID 5 (CKG) in Chuck units-"cut CKG into equal small chunks of Extend-" select Extend to form LUN, map to the host (this is RAID 2.0)

RAID 2.0 because the unit of RIAD is a large Chunk, when a hard disk fails, the hard drives involved in the MINI RAID group (CKG) related to the hard disk are all involved in restructuring, the same amount of data, more people working, it must be soon.

In a word, if you do RAID based on hard disk, it is RAID 1.0.If you do RAID based on a large chunk of Chunk of hard disk, it is RAID 2.0.

As for IBM XIV, he does not do RAID, cuts all the hard drives to 1MB size, and uses pseudo-random algorithms to keep 2 copies between different nodes (a bit like RAID 10), so the hard disk failure recovery time and performance are the same as RAID 2.0, but the capacity utilization is only 50% at most, so we call it "RAID 2.0 -". As for why Huawei is called "RAID 2.0 +", it is said that there are a lot of value-added features on RAID 2.0, and I don't know if my explanation is correct. Of course, Huawei's RAID 2.0 is smaller than 3PAR's cut hard disk block Chunk, so it is more flexible and randomly distributed.

Is it helpful for you to read the above content? If you want to know more about the relevant knowledge or read more related articles, please follow the industry information channel, thank you for your support.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.