No one wants to think about a disaster crippling or even destroying their data center. But even as hurricane season has ended for Atlantic and Gulf Coast states, wildfires are raging in Southern California. Earthquakes are an ever-present danger. Disaster planning is moving higher up in the priority list for many data center managers.
Disaster recovery (DR) planning typically focuses on data protection and application availability. Most organizations consider the information maintained on servers and storage devices to be infinitely more valuable than the technology itself. However, DR plans should also include provisions for protecting equipment from physical damage.
Location : Ideally, data centers would be located in a geographical area that's not prone to natural disaster. That's seldom possible, so organizations must do the best they can to isolate it from any disaster that does occur. That means locating the room in an interior room or at least as far away from windows as possible. In areas where hurricanes and tornadoes are the greatest threat, an underground location may be the best option (without flooding is a problem). In earthquake zones, it's critical to select a well-constructed building that's compatible with the latest codes.
Backup Power : Power outages are a leading cause of equipment downtime, and UPS failure is the No. 1. 1 cause of unplanned equipment outages. UPS should be carefully selected, implemented and maintained to ensure a steady supply of conditioned power with a regulated voltage level.
Fire Suppression : Many data centers rely on conventional sprinkler systems, but water can destroy equipment and cause other problems as well. A better approach is to employ a dry "pre-action" system that will extinguish most fires before the sprinkler system is activated. Modern fire-suppression systems use halo-carbons, which remove heat from fires, or inert gases, which deprive them of oxygen. Both can provide excellent fire suppression if the system is properly designed, installed and tested. The fire alarm should also be tested – if it is faulty, the fire-suppression system might not be activated.
Flood Control : If the data room is located in a flood-prone area, a dumping system should be installed. The system should activate automatically and be connected to generator power so that it lasts operation if the electric grid goes down.
Earthquake Protection : In earthquake-prone areas, it's important to select racks and cabinets that are rated to withstand seismic activity. These units typically have special mounting brackets to attach them securely to the floor.
Flexible Processes : Data center personnel should understand their responsibilities and be thoroughly trained in DR procedures. Equipment should be monitored by at least one person at all times. Run-books should be kept up-to-date so that equipment can be recovered or reconfigured quickly in an emergency. DR processes should also be well-documented but flexibility is important. Personnel should feel empowered to make decisions and improvement based on the situation at hand.
Test : In most organizations, the DR plan is seldom, if ever tested. The plan should be tested at least twice a year and updated as the environment and business priorities change.
These 7 steps can help build design a flexible and resilient data center infrastructure and select systems that will protect your valuable equipment.