|
|
![]() | |
|
|
|
To access the contents, click the chapter and section titles.
Sams Teach Yourself MCSE Windows NT Server 4 in 14 Days
6.3. Configuring RAID1RAID1 is implemented in NT Server 4.0 as either disk mirroring or disk duplexing. Two disks work in a pair to provide redundancy. All data written to a mirrored partition on one disk is simultaneously written to the other. The original Berkeley RAID specifications outlined the mirroring of entire disks, hence the term disk mirroring. The actual implementation is really partition or volume mirroring. Both disk mirroring and duplexing are considered expensive in terms of overhead. Because all data written to one partition is written to the mirror, there is a 50 percent loss of available space. Mirroring two 100MB partitions yields an available space of 100MB.
RAID1 results in increased read performance because data is read from both disks but generally results in a decrease in write performance because all data is written twice. 6.3.1. Disk MirroringDisk mirroring is really just partition mirroring. Any NT partition can be used in a mirror set regardless of the file system. This includes the boot and system partitions. The most common areas to mirror are the system and boot partitions. This is important because neither the system nor boot partition can be members of a fault-tolerant stripe set with parity. All data is written to both partitions in a mirror set. Should one of the disks fail, the system will continue to run, providing the data from the remaining member of the set. The users will be unaware of the disk failure. The administrator will see the failure noted in the event log. Mirroring requires two physical disks and a single controller as shown in Figure 6.1.
The controller is the single point of failure in disk mirroring. If the controller fails, neither member of the mirror set will be accessible. 6.3.2. Establishing a Mirror SetFollow these steps to establish a mirror set:
6.3.3. Disk DuplexingDisk duplexing is really just disk mirroring with one special difference: the physical disks are each attached to a separate controller. Disk duplexing requires two physical disks and two separate disk controllers, as shown in Figure 6.4.
The addition of the second controller will reduce bus traffic, generally increasing the performance over that of mirroring alone. As with mirroring there is a cost of 50 percent of the disk space used. 6.4. Configuring RAID5RAID5 is implemented on NT Server in the form of stripe sets with parity. It is the with parity that differentiates this implementation from simple RAID0 or stripe sets. A stripe set with parity is created from areas of free space from 3 to 32 physical disks. The greater the number of disks included in the array, the greater the performance and less the overall cost. In order to provide redundancy, parity information is spread across all disks in the array. In earlier RAID implementations such as RAID4, a single disk was used to hold the parity information. If a disk in the array fails, the data can be regenerated from parity information. Overall, the performance of stripe sets with parity is greater than that of mirror sets. If a disk fails, however, read performance will be degraded because the data must be regenerated. The overall space attributed to the parity stripe is equal to the space of one partition in the stripe. For example, a stripe set with parity created from equal 100MB areas of space from three drives would yield a total usable area of 200MB. If a total of five disks were used, the usable space would equal 400MB. 6.4.1. Creating a Stripe Set with ParityThe stripe set with parity is created from areas of free space on a minimum of three physical disks.
|
|
Products | Contact Us | About Us | Privacy | Ad Info | Home
Use of this site is subject to certain Terms & Conditions, Copyright © 1996-2000 EarthWeb Inc. All rights reserved. Reproduction whole or in part in any form or medium without express written permission of EarthWeb is prohibited. |