Disaster Recovery: Los Angeles, CA
Each datacenter is supplied with medium voltage electrical power from the local utility company. Where possible for some of the newer data centers, two independent utility sources are in place, originating from independent feeders or substations. Each datacenter is powered by a dedicated utility step-down transformer for each service. The incoming service is connected to an automatic transfer switch, which is also connected to redundant standby diesel generators. Electrical loads served by the incoming service and generator sources include mission-critical, life safety, HVAC and general-purpose loads.
The mission critical electrical loads at each datacenter are sourced by a UPS (uninterruptible power supply) system. Depending on the datacenter, the UPS in place may be one single source of power (one UPS system), parallel UPS systems which service separate parts of the data center, or redundant UPS systems. Redundant UPS systems or Hi-Tech systems, which are configured with automatic static-bypass and manually operated full-maintenance bypass circuits. The primary UPS system consists of on-line, stand-alone modules providing conditioned, uninterruptible power to critical electrical loads. Some datacenters have distributed redundancy achieved through a reserve UPS system. Where it’s feasible to be implemented by the datacenter, customers may operate on redundant power source as an add-on cost to be incurred by the customer.
Whenever possible, these locations makes use of PMMs and/or PDUs on raised floors to provide for a physically integrated and electrically redundant system for source selection, isolation, distribution, monitoring and control of power to internal and customer computer loads.
Each location also has diesel engine generators in place to provide power to all critical equipment and customer loads. Generators may be located indoors or outdoors depending on site-specific conditions. Separately installed main fuel tanks provide a source of fuel to all engine-generators. Base tanks or “day tanks” are capable of providing up to 15,000 gallons of fuel storage. These on site fuel storages are sufficient to provide at least 12 hours of design load operation (or as much fuel as local authorities will permit). Each datacenter also has a short-notice refueling contract for the diesel generators to ensure sufficient generator back-up capabilities.
Facilities are equipped with fire extinguishers throughout the premise. Dry chemical or clean agent extinguishers are installed in the mission critical space, or adjacent areas, where one might reasonably expect a person to carry them into the affected areas during an emergency.
The fire suppression system is monitored 24×7 by an external alarm monitoring company. The monitoring company will dispatch the city fire department upon receipt of an alarm. Inside some of the datacenters, software is used for fire detection and monitoring, combined with customized floor plan graphics to illustrate detection devices and fire zones to aid data center staff and the fire department in responding to and coordinating all fire control activities.
Sprinkler systems in the datacenters are installed with double interlock pre-action and detection systems functionality. The systems are designed in such a manner that water does not enter the sprinkler system piping during normal operations. Pre-action detection and intelligent heat detectors are installed in the ceiling of mission critical areas. Upon activation of any of these heat detectors, audio-visual alarms (such as horn and strobe lights) will activate throughout the space affected. A signal will be sent to a pre-action valve for the affected zone. If the temperature in the at-risk area also reaches levels to melt any of the sprinkler head fusible links, water is triggered to enter the sprinkler pipes for the affected areas within the datacenter.
Estimated Restore Times For Full Data Loss
Because each server (even at the same location) can vary in disk usage, it’s unfortunately not possible to give exact estimates on how long a full restore will take. From experience however, we have found that most servers can be restored within 8 to 16 hours.
For information regarding restoring VPS services, please click here.