A Primer on Disk Drive Configurations

Redundant Array of Inexpensive (or Independent) Disks (RAID)

A method of using a series of hard disks as an array of drives. Some RAID implementations improve performance. Others improve performance and provide fault tolerance.

fault tolerance

The use of hardware and software to prevent the loss of data in the event of a system, hardware, or power disruption or failure.

Several disk drive configurations are available. You can create the configurations listed next by using software and hardware. Many of these types of configurations are called Redundant Array of Inexpensive (or Independent) Disks (RAID). RAID storage systems require two or more disks, depending on the configuration. Each configuration has a specific level identifier that indicates the type of configuration. RAID systems are critical for servers because they provide improved performance, and in some cases fault tolerance.

Volume Set  A volume set extends the size of a partition beyond a single physical drive. This is not a RAID configuration but is often the result of a RAID configuration. Disks that are used in a RAID configuration are physically separate from the RAID controller but appear as one large drive to software such as Windows 2000, Windows XP, or Windows Server 2003.

Disk Stripe Set-RAID Level 0  A disk stripe set combines several logical partitions of the same size into a single logical disk. Data is striped evenly over each partition. Performance is excellent in this RAID configuration, because of simultaneous reads and writes to all drives in the stripe; however, there is no way to recover data in the event of a hard disk failure.

RAID Level 0 is a good choice for digital-video editors and digital animators who need the storage capacity of many hard drives without sacrificing performance. Because of the improved performance inherent in RAID 0, you might consider it an option for disk-intensive applications such as databases. But keep in mind that RAID 0 provides no fault tolerance. If a disk fails, all data is lost on all drives!

Note 

On most computers that you purchase, the operating system and other software are already installed. The hard drive is usually represented by a single letter, which typically is C. Rarely, if ever, are other configurations present, regardless of the size of the drive. But there are other configurations to consider that can add better performance or higher insurance that the data will be protected.

Mirrored Set-RAID Level 1  A mirrored set contains a primary partition and a secondary partition. Anytime data is written to the first partition, it is automatically written to the second partition. RAID Level 1 is often used to protect the drive where the operating system is located. The mirrored set provides the added fault tolerance needed for servers. It is not typically recommended that mirrored sets be created for all drives. Mirrored sets cost twice as much as non-RAID or RAID 0 to implement because for each disk you want to partition, you need a second disk of the exact same size.

parity

In the context of a stripe set, a series of mathematical calculations based on the data stored. If a disk fails, the stored parity information can be used to rebuild the data.

Stripe Set with Parity Drive-RAID Level 3  A stripe set with a parity drive is similar to a stripe set with a parity stripe. In a stripe set with a parity drive, the parity information is stored on a single drive as opposed to being striped across all drives.

Stripe Set with Parity Stripe-RAID Level 5  A stripe set with a parity stripe is similar to a stripe set, but it contains a parity stripe across all drives. This gives you the benefits of a stripe set while also offering fault tolerance. If one of the drives fails, the parity information from the other drives can be used to rebuild the data onto a new drive. RAID Level 5 is the most common configuration because of the low cost, ease of recovery of a failed disk, and performance. In cases where data availability is critical, such as databases, RAID Level 5 is the preferred choice. It adds a high level of reliability, provides good performance, and is relatively reasonable in cost.

Tip 

When deciding on a disk drive configuration, always consult the manufacturer of the computer first. Not only might you get some free instruction on the best type of configuration for your needs, you might even get some help implementing your new configuration.

Volume Set

A volume set extends the size of a single partition. For example, assume that you created a single partition on your 1GB physical drive. You used the partition to store a database that is approaching 1GB. You can add a second physical drive and with the free space create a volume set.

In volume sets, the data is written sequentially, so you can extend volume sets at any time without having to back up the data, create the volume set, and then restore the data. This makes volume sets an easy way to quickly and temporarily handle a shortage of disk space until you replace your disk drive with a larger drive.

Pros and Cons

Consider these factors when using volume sets:

Same-size partitions?	No. With volume sets, the partitions within the volume set do not have to be the same size.
Performance increase?	No. Volume sets write data sequentially, so there is no performance increase.
Fault tolerance?	No. Because volume sets contain no parity information, they are not fault tolerant. This means that if any physical drive within the volume set fails, the entire volume set is unusable.

Note 

With Windows NT, Windows 2000, Windows XP, and Windows Server 2003, you can have up to 32 disks within a volume set or spanned volume, respectively.

Disk Stripe Set

In a disk stripe set, or RAID Level 0, you define logical partitions of the same size as a stripe set. There must be at least two partitions participating in a stripe set. After you create a stripe set, data is written across the set in stripes. The benefit of this disk configuration is that it enables you to take advantage of multiple disk I/O channels for improved performance.

Input/Output (I/O) channel

A circuit that provides a path for an input or output device to communicate with the processor.

Pros and Cons

Consider these factors when using stripe sets:

Same-size partitions?	Yes. Logical partitions have to be the same size in a stripe set.
Performance increase?	Yes. If the stripe set is located on multiple I/O channels, you will see a performance increase.
Fault tolerance?	No. Because stripe sets contain no parity information, if any drive within the stripe set fails, the entire stripe set will be lost. In this case, you would restore your data from your most recent backup.

Tip 

If you want fault tolerance, it is technically possible to mirror a stripe set. The stripe set must be created using special RAID hardware. When Windows boots up, the stripe set appears as a single logical drive and not as multiple physical disks even though the latter is true. Using the Disk Administrator in Windows NT or Disk Management software in Windows 2000, XP, or Windows Server 2003, you create a software-based mirror set of two striped sets. The problem with software-based mirrored drives is that they are much slower than hardware-based drives. Also, software-based RAID configurations are prone to errors, which could cause data loss. It is not possible to mirror two stripe sets that were created with software.

Mirrored Set

A mirrored set, RAID Level 1, consists of a primary drive and a secondary drive. Anytime data is written to the primary drive, it is copied (or mirrored) to the secondary drive. The benefit of a mirrored set is that if a disk fails, you do not lose any data.

There are two types of mirrored sets:

Disk mirroring  Uses one controller with two disks

Disk duplexing  Uses two controllers and two disks

Pros and Cons

Consider these factors when using mirrored sets:

Same-size partitions?	Yes. In a mirrored set, both logical partitions have to be the same size.
Performance increase?	Yes and no. With disk mirroring, performance decreases on writes, because one controller must write to both drives. With disk duplexing, you see a slight performance increase, because you are using two separate I/O channels.
Fault tolerance?	Yes. Mirrored sets are fault tolerant.

Stripe Set with Parity Drive

A stripe set with a parity drive, RAID Level 3, arranges logical drives of equal size into a stripe set. A separate drive stores parity information. In the event that a data drive fails, the parity information and the data on the functioning drives are used in a mathematical calculation that can reconstruct the lost data.

Pros and Cons

Consider these factors when using stripe sets with a parity drive:

Same-size partitions?	Yes. Within any stripe set, partitions must be equal in size.
Performance increase?	Yes. Each I/O channel that the stripe set uses increases the performance of the stripe set.
Fault tolerance?	Yes and no. As long as only one drive fails (assuming that the parity drive isn't the failed drive), the stripe set with the parity drive is fault tolerant. If two or more drives fail, you must re-create the stripe set with the parity drive and then restore the data from the most recent backup.

Note 

Although most RAID configurations are implemented using special RAID hardware, RAID 3 is not typically available as a software-based option. Windows NT Server 4, Windows 2000 Server, and Windows Server 2003, for example, only support RAID 0, 1, and 5.

Stripe Set with Parity Stripe

A stripe set with a parity stripe, RAID 5, arranges logical drives of equal size into a stripe set. Each drive within the set has a parity stripe. As with stripe sets with a parity drive, discussed in the preceding section, the parity information and data from the functioning disks can be used to re-create a failed disk. Unlike the previous RAID examples, RAID 5 requires a minimum of 3 disks with a maximum limitation of 32 disks.

Pros and Cons

Consider these factors when using stripe sets with a parity stripe:

Same-size partitions?	Yes. Within any stripe set, partitions must be equal in size.
Performance increase?	Yes. Each I/O channel that the stripe set uses increases the performance of the stripe set.
Fault tolerance?	Yes and no. As long as only one drive fails, the stripe set with the parity stripe is fault tolerant. If two or more drives fail, you must re-create the stripe set with the parity stripe and then restore the data from the most recent backup.

Note 

RAID 5 is not supported on Windows 2000 Server or Windows Server 2003 in the same way that it is on Windows NT Server 4. Windows 2000 Server and Windows Server 2003 distinguish between basic disks and dynamic disks. Basic disks have the same capabilities as hard disks did on Windows NT with the exception of RAID 5. Windows 2000 Server introduced dynamic disks that do support RAID 5. Dynamic disks minimize the need to restart Windows 2000 Server or Windows Server 2003 when a drive change such as creating a RAID 5 volume or mirror set has occurred. The RAID 5 drive configuration is supported by Windows NT and is considered a software implementation of RAID 5. With software implementations of RAID, you can use any disk drives that are of equal size.