The various existing RAID levels

The principle behind RAID is to aggregate in one way or another multiple hard drives to join them in a single volume. what are the different ways to create a RAID that we will discuss today. With the program, the main principles, but especially the practice, with the differences between hardware and software implementations. Indications that I think helpful to those considering a NAS, among others.

RAID, kézako? (Theory)
This is an acronym for Redundant Array of Independent Disks (or Inexpensive Disks, which was much more true at a time) can be translated as “redundant array of independent disks “. It is therefore subject to aggregate multiple disks in different ways according to the purpose of the created storage: increased performance, fault tolerance, abstraction of such a storage …

Simple Aggregation (or JBOD)
JBOD stands for just a Bunch of disks, ie “just a set of disks.” It is simply to add the following discs each other, and “merge” disk space seen by the operating system. If one drive dies, one loses only the data that was written on it (a bit more when we did toy with NTFS that is too stupid to avoid fragmentation).

stripping, or RAID0
the RAID0 mode is intended to increase performance in reading, writing, “parallelizing” the discs on the operations of the “cluster”. The disk space is available then, as JBOD, the addition of one of the discs. Big drawback, if one drive falls, we lose the entire volume, and thus data, as they are cut and spread on each disk to be read in parallel.

mirroring or RAID 1
Do I really need you to translate the term mirroring? Typically used with two discs, this mode does is manage the mirrored disks: completeness of data is read and written to both drives. The big downside is that we only see the disk space of the smallest of the cluster units. The big advantage is that if one of them dies, the data is intact, since these duplicate. There is therefore a greater fault tolerance.

This is generally the way we use in the NAS carrying two hard drives, although presented earlier are also available, as well as dedicated servers, often left configured in RAID1 by default when they contain two discs.

Aggregation striped with parity distributed, or RAID5
This is as violent description, but it fits well. It is more difficult to understand, certainly, but in the end, we realize that this is powerful. This mode is used with a minimum of three disks. The data is divided into two parts, written in parallel on two discs, and parity is calculated and written in the third, and it runs continuously (each disc hosts a parity in turn block). The idea is that if one drive dies, although we have actually lost some data, they can be recalculated from the remaining data and parity are available.

Space disk perceived corresponds to all least one disc (since consumes the space of a disk to store parity), two disks to a cluster of three. Performance is also increased compared to a single disk, in the end what is, on paper, the best of all worlds: better fault tolerance, as the RAID1, and faster volume, as in RAID0. In practice, the calculation cost can erase the performance aspect following the block size selected.

The other modes
These modes are the most widespread, and widely used. There exist others, with more or less interest, such as RAID6, which works on the same principle as RAID5, but with two drives for parity instead of one, thus increasing power needs necessary calculation. Or RAID10 (or RAID01), making mixed modes 0 and 1 (eg, RAID0 mirror, or a mirror RAID0). Expensive hard drives, they were ultimately of little interest.

The various implementations of these principles
of the high-end storage systems, and on some workstations, RAID management of these modes is entrusted to specialized storage controllers have their own controls for management, which is done before the startup of the operating system; they are completely independent. For our old PCs, servers or deskside, are often found in the form of expansion card PCI-Express (yes, the same as graphics cards).

On viewpoint of the OS, it sees only one “disk”, this solution makes it compatible with all systems, without necessarily require specialized drivers, access may be through the commonly used protocols, SATA on your mind. Command line tools to query the controllers are often available and still needed to know in real time the disk’s health (and to some extent working on the RAID without having to cut the machine).

These controllers are halfway between the hardware and the software (we will address just after). In practice, if part of the “wiring” hardware is present, always a software and / or driver dedicated to the operating system to manage. They are found almost exclusively integrated into motherboards, and usually only drivers for Windows are available, making the little portable solution.

software RAID
Again, we must distinguish two cases, which stands at the core of the operating system or file system.

operating system

Notably Linux infrastructure md defines a RAID array regardless of the file system being installed. We can now create volumes from complete disks or partitions on these disks. The stress then remains the same: the cluster size equal to that of the smallest volume, or a multiple thereof in the case of RAID5 or 6.

File System

Indeed, aficionados of Solaris (and some BSD) will tell you, nothing beats ZFS. This file system has special strengths, including the ability to precisely manage RAID volumes on its own level (RAID-Z mode). It can implement most modes defined at the beginning of the note, and adds its own features and benefits.

The disadvantages? His license restricted to Solaris, now owned by Oracle, and some BSD shot (it is especially useful in FreeNAS, a distribution to build your own … NAS). Although some crafts (based particularly FUSE) allow to have access to Linux. A module exists, but it is kept outside the main branch of noyal. Morality, there is a btrfs (pronounced better FS) which wants a native replacement for the Linux kernel. Still young, it can also aggregate volume under different conditions. It is largely developed by Oracle, which for once playing good student when the development of free software.

Windows also knows how RAID
apart from hardware RAID, which ultimately abstract storage, Windows also knows how to create volumes in different modes. Personally I have never experienced (and as is Windows, I avoid the counselor because nothing says that it works with Windows to another). So I would be hard to explain in what way it works (operating system or file system, Microsoft virtually working with NTFS). So rather ask the question of Julien Computerz.Solutions , he certainly knows more than me about the Microsoft universe)

There’s other methods to manage its volumes
If you want more details on what I have told about the RAID, you can turn to the section of Wikipedia kivabien is very well done and in French (for . once …)

outside the JBOD and RAID0, all RAID modes are at least the same goal: increase the fault tolerance of the storage system of a machine. If this notion you matter, another mechanism duster storage be interested: LVM. But that’s for another day.

No comments:

Powered by Blogger.