A: RAID is a way of combining multiple disk drives into a single entity to improve performance and/or reliability. There are a variety of different types and implementations of RAID, each with its own advantages and disadvantages. For example, by putting a copy of the same data on two disks (called disk mirroring, or RAID level 1), read performance can be improved by reading alternately from each disk in the mirror. On average, each disk is less busy, as it is handling only 1/2 the reads (for two disks), or 1/3 (for three disks), etc. In addition, a mirror can improve reliability: if one disk fails, the other disk(s) have a copy of the data. Different ways of combining the disks into one, referred to as RAID levels, can provide greater storage efficiency than simple mirroring, or can alter latency (access-time) performance, or throughput (transfer rate) performance, for reading or writing, while still retaining redundancy that is useful for guarding against failures.
Although RAID can protect against disk failure, it does not protect against operator and administrator (human) error, or against loss due to programming bugs (possibly due to bugs in the RAID software itself). The net abounds with tragic tales of system administrators who have bungled a RAID installation, and have lost all of their data. RAID is not a substitute for frequent, regularly scheduled backup.
RAID can be implemented in hardware, in the form of special disk controllers, or in software, as a kernel module that is layered in between the low-level disk driver, and the file system which sits above it. RAID hardware is always a "disk controller", that is, a device to which one can cable up the disk drives. Usually it comes in the form of an adapter card that will plug into a ISA/EISA/PCI/S-Bus/MicroChannel slot. However, some RAID controllers are in the form of a box that connects into the cable in between the usual system disk controller, and the disk drives. Small ones may fit into a drive bay; large ones may be built into a storage cabinet with its own drive bays and power supply. The latest RAID hardware used with the latest & fastest CPU will usually provide the best overall performance, although at a significant price. This is because most RAID controllers come with on-board DSP's and memory cache that can off-load a considerable amount of processing from the main CPU, as well as allow high transfer rates into the large controller cache. Old RAID hardware can act as a "de-accelerator" when used with newer CPU's: yesterday's fancy DSP and cache can act as a bottleneck, and it's performance is often beaten by pure-software RAID and new but otherwise plain, run-of-the-mill disk controllers. RAID hardware can offer an advantage over pure-software RAID, if it can makes use of disk-spindle synchronization and its knowledge of the disk-platter position with regard to the disk head, and the desired disk-block. However, most modern (low-cost) disk drives do not offer this information and level of control anyway, and thus, most RAID hardware does not take advantage of it. RAID hardware is usually not compatible across different brands, makes and models: if a RAID controller fails, it must be replaced by another controller of the same type. As of this writing (June 1998), a broad variety of hardware controllers will operate under Linux; however, none of them currently come with configuration and management utilities that run under Linux.
Software-RAID is a set of kernel modules, together with management utilities that implement RAID purely in software, and require no extraordinary hardware. The Linux RAID subsystem is implemented as a layer in the kernel that sits above the low-level disk drivers (for IDE, SCSI and Paraport drives), and the block-device interface. The filesystem, be it ext2fs, DOS-FAT, or other, sits above the block-device interface. Software-RAID, by its very software nature, tends to be more flexible than a hardware solution. The downside is that it of course requires more CPU cycles and power to run well than a comparable hardware system. Of course, the cost can't be beat. Software RAID has one further important distinguishing feature: it operates on a partition-by-partition basis, where a number of individual disk partitions are ganged together to create a RAID partition. This is in contrast to most hardware RAID solutions, which gang together entire disk drives into an array. With hardware, the fact that there is a RAID array is transparent to the operating system, which tends to simplify management. With software, there are far more configuration options and choices, tending to complicate matters.
As of this writing (June 1998), the administration of RAID under Linux is far from trivial, and is best attempted by experienced system administrators. The theory of operation is complex. The system tools require modification to startup scripts. And recovery from disk failure is non-trivial, and prone to human error. RAID is not for the novice, and any benefits it may bring to reliability and performance can be easily outweighed by the extra complexity. Indeed, modern disk drives are incredibly reliable and modern CPU's and controllers are quite powerful. You might more easily obtain the desired reliability and performance levels by purchasing higher-quality and/or faster hardware.
A: The different RAID levels have different performance, redundancy, storage capacity, reliability and cost characteristics. Most, but not all levels of RAID offer redundancy against disk failure. Of those that offer redundancy, RAID-1 and RAID-5 are the most popular. RAID-1 offers better performance, while RAID-5 provides for more efficient use of the available storage space. However, tuning for performance is an entirely different matter, as performance depends strongly on a large variety of factors, from the type of application, to the sizes of stripes, blocks, and files. The more difficult aspects of performance tuning are deferred to a later section of this HOWTO.
The following describes the different RAID levels in the context of the Linux software RAID implementation.
- RAID-linear is a simple concatenation of partitions to create a larger virtual partition. It is handy if you have a number small drives, and wish to create a single, large partition. This concatenation offers no redundancy, and in fact decreases the overall reliability: if any one disk fails, the combined partition will fail.
- RAID-1 is also referred to as "mirroring". Two (or more) partitions, all of the same size, each store an exact copy of all data, disk-block by disk-block. Mirroring gives strong protection against disk failure: if one disk fails, there is another with the an exact copy of the same data. Mirroring can also help improve performance in I/O-laden systems, as read requests can be divided up between several disks. Unfortunately, mirroring is also the least efficient in terms of storage: two mirrored partitions can store no more data than a single partition.
- Striping is the underlying concept behind all of the other RAID levels. A stripe is a contiguous sequence of disk blocks. A stripe may be as short as a single disk block, or may consist of thousands. The RAID drivers split up their component disk partitions into stripes; the different RAID levels differ in how they organize the stripes, and what data they put in them. The interplay between the size of the stripes, the typical size of files in the file system, and their location on the disk is what determines the overall performance of the RAID subsystem.
- RAID-0 is much like RAID-linear, except that the component partitions are divided into stripes and then interleaved. Like RAID-linear, the result is a single larger virtual partition. Also like RAID-linear, it offers no redundancy, and therefore decreases overall reliability: a single disk failure will knock out the whole thing. RAID-0 is often claimed to improve performance over the simpler RAID-linear. However, this may or may not be true, depending on the characteristics to the file system, the typical size of the file as compared to the size of the stripe, and the type of workload. The
ext2fsfile system already scatters files throughout a partition, in an effort to minimize fragmentation. Thus, at the simplest level, any given access may go to one of several disks, and thus, the interleaving of stripes across multiple disks offers no apparent additional advantage. However, there are performance differences, and they are data, workload, and stripe-size dependent.
- RAID-4 interleaves stripes like RAID-0, but it requires an additional partition to store parity information. The parity is used to offer redundancy: if any one of the disks fail, the data on the remaining disks can be used to reconstruct the data that was on the failed disk. Given N data disks, and one parity disk, the parity stripe is computed by taking one stripe from each of the data disks, and XOR'ing them together. Thus, the storage capacity of a an (N+1)-disk RAID-4 array is N, which is a lot better than mirroring (N+1) drives, and is almost as good as a RAID-0 setup for large N. Note that for N=1, where there is one data drive, and one parity drive, RAID-4 is a lot like mirroring, in that each of the two disks is a copy of each other. However, RAID-4 does NOT offer the read-performance of mirroring, and offers considerably degraded write performance. In brief, this is because updating the parity requires a read of the old parity, before the new parity can be calculated and written out. In an environment with lots of writes, the parity disk can become a bottleneck, as each write must access the parity disk.
- RAID-5 avoids the write-bottleneck of RAID-4 by alternately storing the parity stripe on each of the drives. However, write performance is still not as good as for mirroring, as the parity stripe must still be read and XOR'ed before it is written. Read performance is also not as good as it is for mirroring, as, after all, there is only one copy of the data, not two or more. RAID-5's principle advantage over mirroring is that it offers redundancy and protection against single-drive failure, while offering far more storage capacity when used with three or more drives.
- RAID-2 and RAID-3 are seldom used anymore, and to some degree are have been made obsolete by modern disk technology. RAID-2 is similar to RAID-4, but stores ECC information instead of parity. Since all modern disk drives incorporate ECC under the covers, this offers little additional protection. RAID-2 can offer greater data consistency if power is lost during a write; however, battery backup and a clean shutdown can offer the same benefits. RAID-3 is similar to RAID-4, except that it uses the smallest possible stripe size. As a result, any given read will involve all disks, making overlapping I/O requests difficult/impossible. In order to avoid delay due to rotational latency, RAID-3 requires that all disk drive spindles be synchronized. Most modern disk drives lack spindle-synchronization ability, or, if capable of it, lack the needed connectors, cables, and manufacturer documentation. Neither RAID-2 nor RAID-3 are supported by the Linux Software-RAID drivers.
- Other RAID levels have been defined by various researchers and vendors. Many of these represent the layering of one type of raid on top of another. Some require special hardware, and others are protected by patent. There is no commonly accepted naming scheme for these other levels. Sometime the advantages of these other systems are minor, or at least not apparent until the system is highly stressed. Except for the layering of RAID-1 over RAID-0/linear, Linux Software RAID does not support any of the other variations.