Sams Teach Yourself MCSE Windows NT Server 4 in 14 Days
(Publisher: Macmillan Computer Publishing)
Author(s): David Schaer, et al
ISBN: 0672311283
Publication Date: 12/15/97

Chapter 6
Data Protection: Fault Tolerance and NT Backup

by David Schaer and Theresa Hadden

6.1. Overview

This chapter provides you with the necessary knowledge to configure and maintain a fault-tolerant NT Server 4.0 installation.

6.1.1. Objectives

This chapter gives an overview of RAID technology and describes when and how to implement RAID1 and RAID5 in NT Server 4.0. Additionally, you also learn the methodology for recovering data for a system equipped with RAID technology. The following are the objectives for the Windows NT Server and Enterprise exams:

•  Choose a fault-tolerant method.

•  Choose the appropriate course of action to resolve fault-tolerance failures.

•  Implement an appropriate backup methodology.

6.1.2. Fast Facts

The following list of facts is a concise picture of the information presented in this chapter. It acts as both an overview for the chapter and as a study aid to help you do any last-minute cramming.

•  NT Server 4.0 supports RAID1 (mirroring or duplexing) and RAID5 (stripe sets with parity).

•  Disk mirroring requires a single controller and two physical disks.

•  Disk duplexing requires two separate controllers and two physical disks.

•  The system and boot partitions can be members of a mirror set but not of a RAID5 stripe set with parity.

•  Stripe sets (RAID0) are not the same thing as stripe sets with parity (RAID5).

•  The overhead of mirroring is equal to 1/2 the amount of disk space used.

•  Performance is superior with stripe sets with parity than with mirror sets.

•  A stripe set with parity is created from equal segments of free space from between 3 to 32 physical disks.

•  If a single member of a stripe set with parity fails, the lost data can be regenerated from parity information held on the remaining members.

•  If two or more members of a stripe set fail it is not possible to regenerate the lost data. The data must be restored.

•  Breaking a mirror set does not delete the data.

•  To back up the registry you must select at least one file on the partition on which the registry resides.

6.2. Redundant Arrays of Inexpensive Disks (RAID)

RAID technology was first proposed at the University of Berkeley to rectify the negatives of SLED or Single Large Expensive Drives. There are two specific areas in which the use of multiple drives would be better suited than a single drive: performance and fault tolerance.

Theoretically, a single drive would reach its maximum performance capacity and become a bottleneck before two less-expensive, albeit slower, drives working together to perform the same work. This concept forms the general basis of stripe sets, or RAID0.

The second, and primary area of concern relative to the examination is fault tolerance. Regardless of how fast a drive is performing, if the disk crashes the data is lost. RAID technology can change the “all the eggs in one basket” method by duplicating data across two drives (RAID1) or by spreading data and parity information across from 3 to 32 disks (RAID5). None of the RAID methods circumvent the need to back up the system to tape.

6.2.1. RAID Basics

The exam will test your knowledge of how RAID technology is provided by and implemented in NT Server 4.0. Fault tolerance can be provided by NT Server 4.0 in the form of either RAID1 (mirroring or duplexing) or RAID5 (stripe sets with parity).

RAID implementations can be either hardware- or software-based. Although both NT Workstation and NT Server can take advantage of hardware-level RAID such as those provided by Compaq and Hewlett Packard, software RAID0, RAID1, and RAID5 are available on the NT Server platform but only RAID0 is supported on the NT Workstation platform.

Neither NT Workstation nor Windows 95 can provide software-based fault tolerance.

Hardware RAID solutions offload the management of the RAID array to a special controller. By providing their own processing and caching capability, hardware solutions almost always outperform software RAID. Additionally, if the disk fails, hardware solutions allow the failed disk to be substituted without bringing the system down. Many hardware solutions even offer a hot-spare, a disk that automatically takes the place of a failed disk in the array. The information that was held on the failed disk is calculated from the parity information in the array and written to the hot-spare in the background.

Software RAID is implemented by the operating system. The same processors and memory used to control the operating system are used to manage the RAID array. If a disk fails it is necessary to bring the system down to substitute the disk. NT Server provides software-based RAID.

6.2.2. When RAID Has No Redundancy (RAID0)

Remember that RAID was designed to provide for two primary functions, fault tolerance and performance. RAID0 provides increased performance but not fault tolerance. There is no fault tolerance because there is no redundancy of data. Think of the zero in RAID0 as zero redundancy. RAID0 can be implemented in both NT Server and NT Workstation in the form of stripe sets. Stripe sets are described fully in Chapter 7, “Managing Users and Groups.”

RAID0 is used to provide performance improvements only. A common implementation is using stripe sets to hold read-only databases.