- Home
- Disaster Recovery Planning
- Disaster Planning for Servers
Disaster Planning for Servers
- By Site Administrator
- Published 11/20/2007
- Disaster Recovery Planning
-
Rating:




Server Disaster Recovery Planning
Every enterprise with one or more servers should have a server room that is secured with a combination lock and a reinforced door with a deadbolt. If the room is not windowless, the windows should be barred. The room should have both fire/heat detection and water detection sensors which set off a local alarm and send a signal to an off-premises monitoring facility. At a minimum, it should have fire extinguishers suitable for electrical fires. Enterprises that have hundreds-of-thousands of dollars in equipment in its server room should consider a built-in fire suppression system.
-
Excess heat is, by far, the most commonly reported cause of server downtime and damage. An enterprise should augment its building air conditioning with a room-size air conditioner that kicks-in only when its thermostat shows that the temperature in the room has risen above a specified level, typically 68 degrees. An additional safeguard is available, a thermostat inside any cabinet which has a cooling fan. When a fan fails and the temperature rises, an alarm should be triggered.
-
Water damage is the second-ranking cause of server downtime and damage, although the damage is rarely greater than moderate. There should be no water pipes in the ceiling above the room, or in the walls that enclose it. The server(s) and associated peripheral equipment should be rack-mounted so that up to six inches of standing water will not affect the equipment. -
Power irregularities are the third-ranking cause of server downtime and damage. An UPS (uninterruptible power supply) should be used to protect all servers against surges, spikes, brownouts, and blackouts. The UPS should have a rating which is it least twice the total KVA requirements of the devices it protects. KVA (Kilo Volt Amperes) is a rating that is calculated by multiplying the number of volts by the number of amperes and dividing by 1,000. While a library may not want to operate its servers on battery back-up for an extended period, the UPS should provide power long enough for an orderly shutdown of all servers.

Each server should be configured with a logging tape drive or DVD (write once) so that all information written to disk is also written to that media. Each evening the logging tape should be removed and stored away from the server room and a new tape mounted for database back-up. Overnight, the content of the disk drives should be written to tape. The next morning, the back-up tape should be removed and stored away from the server room and a new tape mounted for logging that day's transactions. It will then be possible to restore all files using the most recent back and logging tapes. Magnetic media can become unstable with repeated use, therefore, seven logging tapes (one for each day of the week) should be used.
An enterprise may choose to do a full back-up only once a week. If so, all of the logging tapes for the week should be saved so that they and the previous weeks back-up tape can be used to restore the files. The logging tapes and the previous week's back-up tape should be stored away from the server room. In a large facility than may be at the opposite end of the building, but for smaller facilities it should be off-site.
At least once per week, a current back-up media should be sent to an off-site storage facility to protect against the loss of the on-site back-up tape.
Enterprises that can afford RAID (Reduced Array of Inexpensive Disks) should configure their servers with them. RAID technology mirrors everything written to one disk on another disk. If a disk fails, the mirroring disk provides access to the information without resorting to the rebuilding of files from the combination of back-up and logging tapes.
The database server for the automated library system should be available only to enterprise staff in the enterprise facility and the vendor of the automated library system.
Spread The Word
Article Series
-
Disaster Planning for Servers
Related Links
Comments




IT & Business Infrastructure



