• No results found

Some operating systems allow you to configure RAID without the need for special hardware. The operating system will come with some utility that will allow you to configure software level RAID, which is known as software RAID. For example, using Disk Management, a tool included with Win- dows 2000, you can implement software RAID through the operating system. The previous section gave you a brief introduction to the different levels of RAID that can be implemented. The following section will describe in more detail the features, benefits, and drawbacks of software level RAID. Throughout the section we will pause and examine the advantages and the disadvantages of most of the common software RAID levels. You will find these highlighted evaluations in shaded sidebars within this section.

RAID 0, also known as disk striping, can be implemented on a server but it does not offer any fault tolerance. So if your servers are hosting mission critical data, this will not be an appropriate solution.

With RAID 0, data is broken down into blocks and written across multi- ple hard disks, which increases performance. Performance is also increased because there is no parity overhead. However, should a disk within the array fail, all data is lost and is only recoverable by restoring from a backup copy.

RAID 0 Advantages

 Improved performance—because data is spread across multiple disks, I/

O performance in increased.

 Minimal hardware—RAID 0 can be implemented on two disks.

 Easy implementation—it is easy to implement and can be done using a

tool that comes with the operating system.

 Zero disk overhead—because there is no parity information, there is no

Because RAID 0 offers no fault tolerance, it should never be used for data that is mission critical.

RAID level 1 is also known as disk mirroring. This is one of the most common implementations of RAID in a server environment. With disk mirroring, two disks are required so that data from one disk can be copied or mirrored onto a second disk. Each time a write is made, it is duplicated to the second disk in the mirrored set. If the first disk fails, the data can be accessed from the second disk in the mirrored set (see Figure 5.2).

F I G U R E 5 . 2 RAID 1

RAID level 1 has little impact in terms of performance. You will not see any increase in performance when reading from the disk and you may see a decrease in performance for disk writes because the data now has to be written to two different disks.

RAID 0 Disadvantages

 No fault tolerance—if a single disk fails, all data is lost and must be

restored from backup.

 Increased single point of failure—most implementations of RAID 0 can

support up to 32 disks. However, the more disks you have, the more points of failure. If any of the disks within the array fail, all data will be lost. Disk 1 C: Disk 2 C:' Disk Controller

In terms of fault tolerance, RAID level 1 can withstand the failure of one disk without data loss. This implementation however cannot withstand the loss of a second disk, so when disk failure occurs, it is important to replace the disk as quickly as possible.

A variation of RAID 1 is disk duplexing. It is similar to disk mirroring but provides an additional level of fault tolerance. With disk mirroring, the disks in the array use the same disk controller. Should the disk controller fail, both disks fail as well. With disk duplexing, each of the hard disks has a separate controller, adding yet another level of fault tolerance.

Some implementations of RAID allow you to add a hot spare, which can be used if one of the disks in the array fails.

The second most common level of RAID is RAID level 5, also known as striping with parity. It requires a minimum of 3 disks and most implementations

RAID 1 Advantages

 Easy to implement fault tolerant options. Since data is duplicated on

two disks, it is readily available should one of the disks fail.

 Disk duplexing—this process can be implemented to eliminate the disk

controller as a single point of failure.

 Fault tolerance for system partition—normally, boot files cannot be

written across multiple disks. RAID 1 allows you to implement fault tolerance for the system partition.

 It offers a relatively low-cost fault tolerant solution because only two

disks are required.

RAID 1 Disadvantage

 Disk overhead costs—the major disadvantage of RAID 1 is the disk

overhead or the cost per megabyte. A large amount of disk space can be consumed mirroring one disk onto another. RAID 1 has the highest disk overhead of any RAID implementation.

support a maximum of 32 disks. Data that is written to the array is stored across multiple disks. This level of RAID provides fault tolerance by adding parity information that can be used to regenerate the data should a disk in the array fail. The volume size for the array is determined by the smallest disk (all disks in the array must be the same size). For example, if you have three disks, 10GB, 20GB, and 40GB in your array, 10GB from each disk will be used because this is the smallest disk size. The total volume for the array would be 30GB, with 20GB available for data storage because 10GB is reserved for parity.

In terms of fault tolerance, RAID level 5 can withstand the loss of a single disk and use the parity information to recover the lost data. If more than one disk in the array fails, all data is lost.

Not only does RAID level 5 provide fault tolerance for data, but it can also increase performance because it is faster to read data from multiple disks than it is from a single disk. You may see a slight decrease in write per- formance because the parity information has to be calculated. The trade-off for fault tolerance is a slight increase in server overhead. A portion of your disk volume will be used to store the parity information. To calculate the amount of disk space for the parity information use the formula 1/x where x is the number of disks in the array. For example, if you have four 20GB disks in your RAID implementation, the total volume is 80GB but 20GB is reserved for parity information (1/4 of the total volume space).

Level 5 Advantages

 Fault tolerance—obviously, one of the main advantages to RAID 5 is

fault tolerance. It can withstand single disk failure while still providing access to the data, using the parity information to re-create the data stored on the failed disk.

 Increased performance—RAID 5 offers an increase in performance

because data, including the parity information, is written across multiple disks.

 Replace without shutting down—depending on your server configura-

tion, if it supports hot swapping, you will be able to replace the failed disk without shutting down the server.

 Low disk overhead—the disk overhead for RAID 5 is considerably lower

Related documents