Backup and Reliability Requirements

One of the most powerful characteristics of Windows NT Server is its reliability. Windows NT Server includes the following reliability and fault-tolerance capabilities:

• Automatic restart
  
  
• Complete tape backup support
  
  
• Error handling and protected subsystems
  
  
• Fault tolerance using disk mirroring, disk duplexing, and disk striping with parity (RAID 5)
  
  
• Recoverable file system, automatic replication of directory partitions, and support for Backup Directory Servers
  
  
• Uninterruptible power supply (UPS) support

The following sections define hardware requirements and recommendations for peripheral devices that support reliability and availability under Windows NT Server.

Backup Hardware

This section defines the requirements for backup hardware for servers.

System includes integrated backup solution

Recommended

An integrated tape drive or other device should be included in the system. Although the recommended method to back up files on another server is to use a backup service under Windows NT Server, system administrators will commonly want offline backup capabilities to be available.

For information about hardware requirements related to backup capabilities, such as tape drives and so on, see Chapter 5, “Storage Device Requirements.”

Power Supply

This section defines the guaranteed power requirements for servers.

System includes UPS provided with system

Recommended

For servers deployed in many corporate environments, the more common choice will be to provide guaranteed power for the server room.

System includes power supply protection using N+1 (extra unit)

The system overvoltage/undervoltage protection and power supply switch-over circuitry should have the capability to regulate according to the system load. For each voltage used in the system, the output voltages of the redundant power supplies should be within the range of values that can guarantee the proper operation of the system, no matter which supply is active. Power-supply switch-over should occur swiftly enough to maintain normal server system operation.

System supports replacement of power supplies

Recommended: Hot-swapping capabilities for power supply replacement and power supply redundancy.

System supports replacement of fans

For highest reliability, hot-swap fans should be implemented to maximize server up time.

System includes local hot-swap power supply replacement indicators

Any system that provides hot-swap power supply replacement and power supply redundancy should have local indicators that unambiguously indicate the supplies that must be replaced. These indicators guide service personnel to replace the correct power supply. Automatic retention mechanisms that prevent incorrect supply removal also satisfy this guideline.

Fault-Tolerant Hardware

This section provides design guidelines for fault-tolerance features and capabilities.

System includes multiple hard drives

Recommended: Hot-swappable drives.

Bus mastering is required for the drive controllers. Use of multiple hard drives and controllers in a server system provides both performance and reliability benefits, therefore, it is recommended for server systems.

System includes intelligent RAID controller with adequate storage capacity

An intelligent RAID controller—where the controller itself has the capability to run the array management software locally rather than simply executing disk accesses for host-based array software—provides the benefit of reduced demands on the host processor or processors, thereby freeing those computing resources and allowing their use by other tasks. The intelligent RAID controller may be internal to the server chassis, or within an external drive enclosure.

If an intelligent RAID controller is provided in a system, it should be capable of handling sufficient amounts of disk storage to fulfill the needs of the targeted usage model for that server. These needs will vary based on the storage-intensive nature of the server’s tasks.

System supports at least one of RAID 1, 5, or 1/0

RAID 1 and RAID 1/0 are recommended. RAID 5 is also acceptable. RAID 0 (for enhanced performance but no added reliability) is optional.

RAID support includes notification of failed drive

Required

If RAID support is implemented, notification of a failed drive must be provided by the disk subsystem, with notification sent to the system administrator.

RAID subsystem supports automatic replacement of failed drive

The RAID subsystem must provide automatic replacement of a failed drive by a standby disk and must rebuild lost data without interfering with system operations.

RAID subsystem supports manual replacement of failed drive

The RAID subsystem should provide for manual replacement of a failed drive without shutting down or halting the system. The subsystem should also allow lost data to be rebuilt without interfering with system operations beyond some decreased performance of drive array access.

Serviceability Requirements

This section provides design guidelines for serviceability features and capabilities.

System includes protected forced NMI switch for system diagnosis

Recommended

The system should include a protected switch to force an NMI on a stalled system. This switch permits the system to perform a memory dump that can then be used for diagnosis of system failures. If implemented, this switch must be protected in such a way that only an authorized administrator can perform this action.