Energy aware processing load distribution system and method

A method for controlling a data center having at least one load balancing switch for receiving and distributing a plurality of requests for performing services by a plurality of server systems in a facility which has at least one cooling system for cooling the plurality of server systems, the respective server systems having at least an active state and a low power quiescent state, the method comprising:

• collecting information from at least one thermal sensor, and hardware devices and software instrumentation maintained within each respective server system, comprising at least one computing utilization activity counter;
• modeling a current state of the data center comprising a processing load on each respective server system, and at least one temperature, and estimating at least a compute performance of at least one server system, based on at least the collected information;
• anticipating trends in the state of the data center based on the history of collected information, and the modeled current system state with at least one automated processor;
• automatically deciding at the end of predetermined time intervals a number and identity of server systems which are to be active over a prospective time interval with an automated processor, in dependence on at least the anticipated trends;
• controlling the at least one cooling system to provide at least sufficient cooling to each respective server system to ensure that an anticipated trend in the at least one temperature does not exceed a predetermined threshold;
• and
• selectively automatically directing requests from the at least one load balancing switch to a subset of the server systems which are to be active, having an available load processing capacity to process the request, the selectively automatically directing being selectively dependent on an anticipated compute performance impact of a respective request.

A system for controlling a data center comprising at least one load balancing switch configured to receive and distribute a plurality of requests for performing services by a plurality of servers in a facility, the respective servers each having at least an active state and a low power quiescent state, the system comprising:

• at least one input port configured to collect information from hardware devices and software instrumentation maintained within each respective server, comprising at least one computing utilization activity counter;
• a memory configured to store information defining a model of a current state of the data center encompassing at least a processing load on each respective server, based on at least the collected information from the hardware devices and the software instrumentation;
• and
• an automated processor configured to anticipate trends in the state of the data center based on a history of collected information, and the model of the current system state of the data center;
• to automatically decide a number and identity of servers which are to be active over a prospective time interval, in dependence on at least the anticipated trends;
• to model at least a compute performance of at least one server;
• and to selectively automatically direct requests from the at least one load balancing switch to a selected subset of the servers which are decided to be active, having an available load processing capacity to process the request, selectively dependent on an anticipated compute performance impact of a respective request.