Comprehensive introduction to RAID disk array technology

In the early days of computer development, the price of "large capacity" hard disk is still quite high, the main way to solve the problem of data storage security is to use devices such as tape drive to backup, although it can guarantee the security of data, the work of lookup and backup is quite tedious. In 1987, three engineers, Patterson, Gibson and Katz, published a paper entitled "A case of redundant array of inexpensive disks (redundant array of Inexpensive disks scheme)" at the University of California, Berkeley, The basic idea is to combine many small, relatively inexpensive hard drives to make them more expensive than a large hard drive. This design idea was quickly accepted, and the raid technology was widely used, and data storage entered a new era of faster, safer and cheaper.

Disk arrays are unfamiliar and mysterious to PC users. The disk array in the impression seems to remain in such a scene: In the spacious hall, lined with disk cabinets, a number of gloomy, early bald engineers wandering among them, and constantly pull out a piece of heavy hard drive, and then insert a piece of seemingly heavier hard drive ... Finally, as the price of high-capacity hard drives continues to drop and PC performance improves, Ide-raid, as the cheapest solution for disk performance improvements, begins to walk into the computer system of the average user.

Introduction of RAID Technical specification

RAID technology mainly includes several specifications, such as RAID 0~raid 7, which have different emphases, and the common specifications are as follows:

RAID 0:raid 0 splits data in bits or bytes sequentially, reads/writes on multiple disks in parallel, and therefore has a high data transfer rate, but it does not have data redundancy and is therefore not a real RAID structure. RAID 0 simply improves performance and does not guarantee data reliability, and one of the disk failures affects all data. Therefore, RAID 0 cannot be applied to situations where data security requirements are high.

RAID 1: It is data redundancy through disk data mirroring, generating data that is backed up on a pair of independent disks. When raw data is busy, the data can be read directly from the mirrored copy, so RAID 1 can improve read performance. RAID 1 is the highest unit cost in a disk array, but provides a high level of data security and availability. When a disk fails, the system can automatically switch to read and write on the mirrored disk without the need to reorganize the failed data.

RAID 0+1: Also known as the raid 10 standard, is actually a combination of RAID 0 and RAID 1 standards, in which the data is continuously split in bits or bytes and read/write in parallel, while disk mirroring for each disk is redundant. The advantage of this is that it has a high speed of RAID 0 and reliable data for RAID 1, but CPU usage is also higher and disk utilization is low.

RAID 2: The data is fragmented across different hard disks, in bits or bytes, and is used to provide error checking and recovery using a coding technique called "aggravating average error correction codes (SEA code)". This coding technique requires multiple disk storage checking and recovery information, making RAID 2 technology more complex to implement and therefore rarely used in a business environment.

RAID 3: It is very similar to Raid 2, where the data is striped across different hard disks, except that RAID 3 uses simple parity and a single block of disk to store parity information. If a disk fails, the parity disk and other data disks can be reused, and if the parity disk fails, the data is not used. RAID 3 provides a good rate of transmission for a large number of contiguous data, but for random data, a parity disk can be a bottleneck for write operations.

RAID 4:raid 4 also bars data and distributes it across different disks, but the bar units are blocks or records. RAID 4 uses a disk as a parity disk, each write operation requires access to the parity disk, then the parity disk will become the bottleneck of the write operation, so raid 4 in the business environment is also rarely used.

RAID 5:raid 5 does not specify parity disks separately, but instead accesses data and parity information across all disks. On RAID 5, the read/write pointer can operate on the array device at the same time, providing higher data traffic. RAID 5 is more suitable for small blocks and randomly read and write data. The main difference between RAID 3 and RAID 5 is that RAID 3 involves all the array disks for each data transfer, while for RAID 5, most data transfers are done on a single disk and can be done in parallel. There is "write loss" in RAID 5, that is, each write will produce four actual read/write operations, of which two read old data and parity information, two times to write new data and parity information.

RAID 6: RAID 6 increases the second independent parity information block compared to RAID 5. Two separate parity systems use different algorithms, the reliability of the data is very high, even if two disks fail at the same time will not affect the use of data. However, RAID 6 requires more disk space to be allocated to parity information and has a much greater "write loss" than RAID 5, so "write performance" is very poor. Poor performance and complex implementations make RAID 6 less practical.

RAID 7: This is a new RAID standard with its own intelligent real-time operating system and software tools for storage management that can be completely independent of the host's CPU resources. RAID 7 can be viewed as a storage computer (Storage Computer), which is significantly different from other RAID standards. In addition to the above criteria (table 1), we can combine a variety of RAID specifications, such as RAID 0+1, to build the desired RAID array, such as RAID 5+3 (RAID 53), which is a more widely used array form. Users can generally gain a disk storage system that is more responsive to their requirements by flexibly configuring the disk array.



Initially, the RAID scheme is mainly for SCSI hard disk system, the system cost is more expensive. In 1993, Highpoint launched its first ide-raid control chip, which makes it possible to build RAID systems using relatively inexpensive IDE hard drives, which greatly reduces the "threshold" of raid. Since then, individual users have begun to pay attention to this technology, since hard drives are the most "slow" and least secure devices in modern PCs, and the data that users store in them often far exceed the price of the computer itself. With relatively low cost, RAID technology enables individual users to enjoy multiple disk speeds and higher data security, and now the Ide-raid control chips in the PC market are mainly from highpoint and promise. In addition, some are from AMI companies (see table 2).

Ide-raid chips for individual users generally provide support for RAID 0, RAID 1, and RAID 0+1 (RAID 10), although they are technically not comparable to commercial systems, but the speed and security assurance provided by the average user is sufficient. With the continuous improvement of the transmission rate of the hard disk interface, the Ide-raid chip has been continuously upgraded, the chip market mainstream chip has all supported the ATA 100 standard, and Highpoint Company's new HPT 372 chip and promise latest PDC20276 chip, You can even support the ATA 133 Standard IDE hard disk. In the motherboard manufacturers increased competition, personal computer users gradually improve today, on the motherboard on the onboard RAID chip manufacturers have a few, users can not purchase a RAID card, directly build their own disk array, feel the speed of the disk hurricane.

Two. Implementation of IDE RAID via hardware control chip

RAID 0 and RAID 1 are the most widely used personal computers in a raid family, and there are few individual users willing to use 4 or more hard disks to build a RAID 0+1 or other hard disk array, so we'll just explain these two raid modes here. We choose to support the Ide-raid function of the Kt7a-r aid motherboard, step-by-step to introduce the installation of Ide-raid. The Kt7a-raid integrates the Highpoint 370 chip, which supports RAID 0, 1, and 0+1.

RAID natural hard drive, RAID 0 and RAID 1 disk requirements are not the same, RAID 1 (Mirror) disk mirroring generally requires two (or more) disk capacity consistent, while RAID 0 (Striping) disk generally do not have this requirement, of course, It is ideal to choose a hard disk with similar performance or even identical capacity. To facilitate testing, we choose two 60GB of Seagate Cool fish Ⅳ Hard drive (Barracuda Ataⅳ, number st360021a).　System Selection Duron 750MHz CPU,2X128MB, GeForce2 Pro video graphics, should be said to be more common configuration, we also want to understand the need to build RAID system requirements. Creation of 1.RAID 0

First step

First, you need to back up the data on your hard disk. Many users do not attach importance to backup this work, especially some more careless individual user. Creating a raid is a risky operation for the data, it's possible to destroy data on the entire hard drive with a little attention, and RAID 0, which we introduced first, is where the data on the disks in all the arrays will be erased, including the hard disk partition table, when the raid 0 o'clock is created. So it's important to have a WINDOWS98 boot disk with FDISK and the format command ready first.

Second Step

Set the jumpers for the two hard drives to master, respectively, with the Kt7a-raid IDE3, IDE4 ports (which are controlled by the HighPoint370 chip on the motherboard). Since RAID 0 rebuilds the partition table of two hard disks, we do not need to consider the order of the hard disk connections (we will see the importance of the order in which RAID 1 o'clock is created).

Third Step

Set up the BIOS to turn on the ATA RAID CONTROLLER. We entered the integrated peripherals option in the BIOS of the upgrade Kt7a-raid motherboard and turned on the ATA100 RAID IDE CONTROLLER. I recommend that the boot sequence of all changes to ATA RAID, we actually found that this is not feasible in the system installation, do not have partitions of hard disk can start? So we still set the floppy drive as a preference.

Fourth Step

The next setup step is to create the core of RAID 0, and we'll show you graphically in detail:

1. Reboot the computer after the system BIOS setup is complete, and the hard drive will not be reported when the boot is detected.
2. The management of the disk will be taken over by the Highpoint 370 chip.
3. The following is a very critical Highpoint 370 bios setting that presses both Ctrl and H at the Highpoint 370 disk scan interface.
4. The first thing to do after entering the Highpoint 370 BIOS Setup screen is to select "Create RAID" for raid creation.
5. For RAID mode selection in "Array mode", you can see the options for RAID 0, RAID 1, RAID 0+1, and span, where we have selected RAID 0.
6.RAID mode selection completion will automatically exit to the previous level of the menu for "Disk drives" option, in general, direct return on the line.
7. The next setting is the stripe unit size, the default value is 64kB, and no special requirements can be ignored. 8. Then is the "start Create (Start)" option, before you press "Y", please seriously think about whether there are important data left on the hard disk, this is your last chance! Once you start to create a raid, all data on your hard disk is purged.
9. After the completion of the creation is to specify the boot boot disk, any choice.

Press the "ESC" key to exit, and of course, press "Y" to confirm.



Highpoint 370 BIOS does not provide a function similar to "Exit without Save", which is irreversible when modified

Fifth Step

After restarting the computer again, we can see the word "Striping (RAID 0) for Array #0" on the screen. Insert the previously made boot disk and start the DOS. Open the Fdisk program, eh? How is a hard drive visible? Yes, the RAID array is already viewed as a hard drive, and for the operating system raid is completely transparent, and we don't have to worry about the management of the raid disk, which is done by the control chip. Next in accordance with the common single hard disk method to partition, you will find that "this" hard disk capacity "changed" big, careful calculation, yes, total capacity is two pieces of hard disk added capacity! We can compare the read and write of RAID 0 to a zipper, which separates data on two hard drives, and reads data faster and does not waste disk space. Don't forget to activate the primary partition after partitioning and formatting.

Sixth step

The choice of operating system is quite a struggle, HighPoint370 chip provides driver support for WINDOWS98/NT/2000/XP, so we chose to support better Windows XP for new hardware, given that the raid feature is targeted at relatively advanced users. Professional English version (the English version of the system is mainly to facilitate the subsequent WinBench test, we use RAID can fully use the Chinese version of the operating system), Windows2000 is also a good choice, But hardware support is clearly not as good as Windows XP Professional.

Seventh Step

For a RAID-based computer, the installation of the operating system is not the same as under normal circumstances, let's take a look at the diagram, which appears after Windows XP completes the first "file copy" reboot, and the installer prompts you to "press F6 to install SCSI devices or RAID disks" in English, which is a very short process. And users tend to ignore the prompts at the bottom of the screen.

The installation selection appears after pressing F6, select "S" will install RAID control chip driver, select "Enter" Do not install.

Pressing the "S" key prompts you to insert the RAID chip drive disk.

Type carriage return, the installer automatically searches for the program on the drive disk, select the "WinXP" and return.

If the version provided is inconsistent with the built-in driver version of Windows XP profesional, the installer prompts the user to make a choice.

Press ' S ' will install the floppy disk provided and press ' Enter ' to install the Windows XP Professional

Self-driven. Press "S" and then need to confirm, this time is to press "Enter" (This ... Confirm too much, hehe). Next is the normal system installation, and the normal installation is no different.

Installation settings for RAID 0 Let's get to the point where we'll talk about the RAID 1 installation. RAID 1 is much simpler to install than RAID 0, and is not disruptive in the right operation.

Creation of 2.RAID 1

Although in principle and RAID 0 completely different, but RAID 1 installation setup process and RAID 0 is not much different, the main difference is the Highpoint 370 bios in the settings. To avoid duplication, we only focus on this part of the setting:

Select "Create RAID" after entering the Highpoint 370 bios:

1. Click Return on "Array mode" and select the second item in the RAID mode selection "Mirror (RAID 1) for data security (create mirror for DataSource disk)".
2. Then the choice of the source disk, we remind the user again: Be careful, do not choose the wrong.
3. Then the choice of the target disk, which is what we call a mirror or backup disk.
4. Then start creating.
5. After the completion of the creation of the BIOS will be prompted to mirror the production, the process is very long.
6. It took us about 45 minutes to complete the 60GB image production, and this RAID 1 creation was completed. RAID 1 copies data from the master disk to the mirror disk, so you need to be very careful when building RAID 1 o'clock, and don't mix the main disk and mirror gossip, otherwise the result will be tragic. RAID 1 can be created on two hard drives with no data, and it can be added on a hard disk that already has an operating system installed, much more convenient than RAID 0 (except for the lengthy mirroring process). After the creation is complete, we try to unplug one of the hard drives, the HighPoint370 BIOS warns, presses "ESC", and the other hard disk takes up the responsibility of the source disk, and all the data is intact.

For adding RAID 1 on a hard disk that already has an operating system installed, our recommended step is to open the control chip in the BIOS → start the operating system install Highpoint 370 drive → shutdown The source and mirror disk in IDE3, 4 → enter highpoint 370 BIOS setup RAID 1 (see above for steps) → Reboot the system to complete creation.

We had a simple test of two types of RAID, although the test scores of RAID 0 were somewhat confusing, but the actual use still felt much faster than a single hard drive, especially when Windows XP Professional started unusually quickly and the progress bar flashed. As for the instability of the transmission curve, we estimate that there is some relationship with the platform selection, after all, the integrated chip is very easy to interfere with this high data throughput work. But even then, we see that the raid 0 system has a very high rate of data transfer, at one point approaching 60mb/s. Compared to RAID 0, the performance of the RAID 1 system has not significantly improved compared to a single disk system, but we found that the work curve of RAID 1 was very stable, with little volatility in the test. Then look at the results of the disk test in WINBENCH99 2.0, at a glance.

For both users and the operating system, RAID 0 and 1 are transparent and do not affect any operation, as we do with a hard disk.



Third, the use of software methods to implement RAID

In addition to using a RAID card or a chip on the motherboard to implement the disk array, we can use the software in some operating systems to implement the RAID function directly, such as the built-in RAID function in WINDOWS2000/XP.

Before understanding WINDOWS2000/XP's software RAID capabilities, let's first look at one of the features in Windows2000-dynamic disk Management.

A dynamic disk, instead of the previous partitioning method, is called a volume set, which is in fact consistent with partitioning, but has the following differences:

1. Can change the disk capacity arbitrarily
Dynamic disks can change the size of the disk without restarting the computer and do not lose data, and basic disks lose all of their data if they want to change the partition capacity (there are also special disk utility software that can change partitions without destroying data, such as Pqmagic).
2. Limits on disk space
Dynamic disks can be extended to discontinuous disk space on a disk, and you can create a volume set that spans disks and combine several disks into one large volume set. The partition of a basic disk must be contiguous space on the same disk, and the maximum capacity of the partition is, of course, the capacity of the disk.
3. Volume set or partition number
The number of volume sets that a dynamic disk can create on one disk is unlimited, and the relative basic disk can only be divided into 4 zones on a single disk, and only one primary and extended partition can be used when using DOS or windows9x.

* Here it is important to note that dynamic disks can only be used in WINDOWSNT/2000/XP systems and other operating systems do not recognize dynamic disks.

Because most users ' disks are basic disk types, in order to use the Software RAID feature, we must convert it to a dynamic disk: Control panel → admin tools → Computer Management → Disk Management, switch one of the Windows to a disk list in the View menu. We can then convert the selected disk to a dynamic disk through the right-click menu.

There are several types of dynamic volumes that can be seen when you divide a dynamic volume.

1. Simple volume: Contains disk space on a single disk, as well as partition functionality.
(when there are two or more dynamic disks in the system and there is unallocated space on two disks, we are able to choose between the following two ways of splitting the volume 2. Spanned volumes: spanned volumes merge unallocated space from multiple disks into a single logical volume.
3. Striped Volume: Combines unallocated space on multiple (2 to 32) disks to a volume.
(We see another partition if the two dynamic disks in the system are consistent as described above)
4. Mirrored Volume: Two copies of the same copy of a single volume, each on one hard drive. That's what we used to say about RAID 1.
When we have three or more dynamic disks, we can use more complex RAID mode to--raid 5, where new volumes will appear in the split-volume interface.
5.RAID 5 Volume: A striped volume that is equivalent to a parity, that is, a RAID 5 mode.
For most PC users, building RAID 0 is the most economical and useful form of array, so we'll just explain the build of software RAID 0 here:
To use software RAID 0 in WINDOWS2000/XP, you must first convert the disks that are ready to be included into the array to the dynamic disks described above (note that you cannot convert basic and dynamic disks in WINDOWS2000/XP's default Disk Management interface, refer to the description above) , where we try to use zoning, which is the difference between software RAID and building a disk array using a RAID chip. We selected a 29GB partition to divide the striped volume, when partitioning a striped volume, the system requires a corresponding partition, which means that there must be an equally 29GB or greater unallocated space on the other dynamic disks, and when the striped volume is allocated, the two volumes of the same size will be merged by the system. At this point, our formatting and other operations are also on two disks.

After building RAID 0, we decided to test its hard drive transfer rate to determine how well the software RAID was performing, and the platform we built the software RAID was not the same as the hardware RAID platform in the previous article, in order to ensure CPU performance to ensure the implementation of our software RAID, We have a higher-end system: Athlon XP 1700+, Samsung 256MB DDR memory, ASUS A7V266-E Motherboard, because the software raid on the hard disk specifications, so the hard disk system we selected a different size of the hard drive, Seagate Cool Fish Ⅳ 60GB and West data 1200BB 120GB two hard drives.

In the second half of the transmission curve, it is clear that the software RAID 0 's hard drive transmission rate is 60mb/s, completely beyond the transmission rate of any one hard drive in the array, and the advantage of RAID 0 begins to manifest. For the high performance users, this should be their dream.

It should be explained here that in the Linux environment, we can also use the Raidtools tool to implement the software RAID function. This tool can produce multiple disk arrays such as soft RAID 0, RAID 1, RAID 4, RAID 5, and so on. Before using Raidtools, first determine whether the Linux core currently in use supports MD. If you are using a core of 2.0.X, and are not compiled yourself, in most cases support soft raid. If you are unsure, you need to compile the core yourself.

While the raid feature gives us a better speed experience and data security, it should be noted that most of the inexpensive Ide-raid solutions available today are essentially "semi soft" raid, which integrates raid control information into RAID chips, Therefore, its CPU occupancy rate is relatively large, and performance is not very stable. This is also why the performance of software RAID 0 in high-end systems can sometimes exceed the "hardware" RAID 0 scenario.



For the user, the High-performance ide-raid storage System, or the need for more robust CPU computing power, or the need for more expensive RAID cards, so the disk array should still be a relatively high-end application. But for beginners, using a simple, inexpensive disk array to improve the usability of your computer's data, or to upgrade your storage speed, is a pretty good choice, and its performance is far from being comparable to high-end systems.

In short, we see a growing number of RAID architectures in the market, especially in the low-end market, the increasing popularity of low-cost Ide-raid solutions and hard disk prices are looking at each other, it seems to indicate the future of personal data storage trends, let's wait

The Highpoint 370 BIOS does not provide a function similar to "Exit without Save", which is irreversible after modification.