|
|
|
| |
        
|
 |
|
To access the contents, click the chapter and section titles.
Sams Teach Yourself MCSE Windows NT Server 4 in 14 Days
6.4.2. Performance Considerations of RAID TechnologyIn the area of performance you must except some basic generalities. The exams do not provide a level of depth that will require advanced calculations to determine which RAID solution is best suited for a given situation. In general, read performance is better on stripe sets with parity than on mirror sets. Reading data from the stripe set will slow if a member of the set fails because the data must be regenerated. Mirrored partitions do provide a modest increase in read performance, but this is countered by slower write performance. The write performance of mirrored sets is considered superior to that of stripe sets with parity. Duplexing provides the best write performance and good read performance.
6.5. Recovering from Hard Disk FailureHard disk failure can occur for many reasons. This section of the exam will test your knowledge of how to recover the system given single- or multiple-disk failures. 6.5.1. When a Mirror Set FailsIf one of the disks in a mirror set fails, the system simply takes the failed disk out of the equation. The system will continue to run without a suspension of services. Because mirror sets commonly include the system or boot partitions, it’s considered good practice to have a fault-tolerant boot disk on hand. The disk essentially takes the place of the system partition. To create the disk, follow these steps:
To restore the mirror you must first “break” the mirror set. Using Disk Administrator, simply select Break Mirror from the Fault Tolerance menu as shown in Figure 6.8. Breaking the mirror set will expose the failed partition, enabling it to be deleted. If an area of free space of sufficient size is available, the mirror set can be reestablished into that area.
6.5.2. When a Stripe Set with Parity FailsIf a single disk in a stripe set with parity fails, the system will continue to function with no loss of data. Of course, because this is an implementation of software RAID5, there could be a significant reduction in system performance until the lost member is recovered. So you will want to regenerate the stripe set as quickly as possible to avoid an excessive amount of system memory and processor time being dedicated to calculating data requests on the fly.
There are two primary causes why a member of a stripe set with parity would fail. The first is a power loss to the specific drive and the second, more severe, is a head crash. Restoring a Single Lost RAID5 MemberWhen a power loss to a single drive causes a member to fail, simply bring the system down, reestablish the proper power, and restart the system. After the system has been restarted with power restored to the member disk, the stripe set with parity can be regenerated using the original members. If the member set was lost due to a disk crash, you must replace the physical disk before you regenerate the stripe set with parity. Restoring a Stripe Set with Parity Following the Loss of Multiple MembersIf multiple members of a stripe set with parity fail at the same time, simply regenerating the stripe set will not be possible. When multiple members fail simultaneously the set will not have sufficient information to regenerate the lost data. The solution is to use a tape backup unit. If multiple disks fail, the stripe set must be re-created—not just regenerated. After the stripe set has been re-created the data that was lost can be restored from tape.
6.6. NT Backup ProceduresBacking up the NT server should be treated as a mandatory task. Even if mirroring, duplexing, or stripe sets with parity have been implemented your system is still susceptible to a system crash. Even if your system never crashes, a user may inadvertently delete or corrupt his data. Because the details of the backup program are stressed on the NT 4.0 Workstation exam (70-63), only the key points of how NTBACKUP functions are covered here. Although it might seem obvious, NTBACKUP was designed to back up to and restore from tape devices only. 6.6.1. Types of BackupsYou can perform five different backup types. The key difference between the backup types is how they handle the archive bit. The archive bit for each file can be toggled on or off (cleared). If it is on, the file is considered to have been modified since the last full backup. NormalA normal backup, sometimes called a full backup, copies all selected files and directories. The archive bit is cleared as the files are backed up. Although this is the longest backup to perform, it will restore the entire system the fastest. CopyA copy is simply that—a copy. The copy option was included to let a user back up a set of files without effecting a standardized backup schedule. The archive bit is not cleared on modified files. DifferentialIf you can think of the D in Differential as standing for Different, this type of backup is easy to remember. All the files that have been modified since the last Normal backup can be selected. The archive bit is not cleared during the backup. If you were to perform a normal backup on Monday and a differential on each successive day, after a crash you must install the normal and then apply the latest differential backup.
|
|
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.
|
)
)
){kind=link}
){kind=link}