Planning a Reliable Configuration
Planning a Reliable Configuration |
|
When choosing a configuration for mass storage, you need to consider variables such as cost, performance, and the time required to completely rebuild a disk. Consider the relative advantages and disadvantages of mirrored volumes, RAID-5 volumes, or a combination of the two techniques.
In general, fault-tolerant configurations are only needed for information that must be readily available in case of hardware failure or unrecoverable disk errors. You do not need to have your page file on a fault-tolerant volume, and it is recommended that you not have the page file on a RAID-5 volume because of potential performance impacts.
You might not need to configure applications on a fault-tolerant volume. If you have applications and other common files on more than one computer running Windows 2000 Server, you can use replication to keep them consistent and to provide the redundancy. If you have applications on a single computer running Windows 2000 Server, you only need to configure them on a fault-tolerant volume if they must always be immediately available. Make sure that you back up the volume every time you install a new application or change default settings for an application.
Note
Placing application data on a striped volume provides the fastest I/O performance for reading data.
If space is a consideration, format your application volume with NTFS (or convert a FAT volume to an NTFS volume), and use NTFS compression for folders and files on the volume.
You need to consider the software and hardware constraints of fault-tolerant storage, as well as the cost and reliability of the system. Each vendor needs to have design guidelines for their system, whether you want a RAID array or want to use the fault-tolerant features provided in Windows 2000 Server.
In certain configurations and situations fault tolerance might not work as you might expect. Check that all of the disk hardware is on the Windows 2000 HCL. If any equipment is not included on the HCL, it might not work well.
If possible, use identical disks. Although this is not a requirement, there are several advantages to using identical hardware for fault-tolerant configurations:
Have a spare disk available and ready for use. When purchasing the disks, it is a good idea to purchase an extra disk for use as a spare. An identical disk guarantees that the spare disk is compatible, does not degrade the system performance, and can be easily installed. If you are using SCSI disks, check the disk documentation to see if you need to manually set the SCSI ID.
Having an available spare or warm spare (a disk that is powered at all times in the configuration and ready to be substituted) allows you to dynamically mirror or replace a member disk of a mirrored or RAID-5 volume in the event of a hardware failure on a spindle.
Have a backup controller available, or configure your disks with duplexed controllers. The system can continue operating when a disk fails, but not when a controller fails. System downtime can be minimized by having a preconfigured controller available. If the replacement controller is not the same as the original, you have to install a new driver, and your configuration becomes more complex.
Note
If you have configured your system volume on a mirrored volume, use the same make and model of controller for your backup, and be sure to use the same translation for both the original and shadow volume.
To a large extent, how you configure your mirrored volumes depends on the number of disks and controllers that you want to have on the computer running Windows 2000 Server. Configuring your boot volume on a disk (and controller) that does not contain data sets results in better performance. There are also other considerations that are related to overall reliability.
When you use a mirrored volume for your system or boot partitions, the following issues must be kept in mind:
Important
Always try starting the computer from both mirrors before you have a failure.
When configuring a RAID-5 volume, buy disks on the basis of cost per megabyte and performance, since storage efficiency increases with a larger number of disks. If you use high performance SCSI controllers, an array of six 9-GB disks provides more efficient storage than nine 6-GB disks. Smaller disks can also be rebuilt faster.
Keep the RAID-5 volume on a different controller and disk than your system and boot volume. Using separate controllers improves performance and can accelerate recovery from hardware failures. Remember that you cannot configure your system volume or your boot volume on a RAID-5 volume.
If the computer itself or the disk containing the boot volume fails (and the boot volume is not mirrored), it might be faster to move the RAID-5 volume to a different computer.