Apparatus and method for efficient estimation of the energy dissipation of processor based systems
A system and method of scheduling tasks, comprising receiving activity and performance data from registers or storage locations maintained by hardware and an operating system; storing calibration coefficients associated with the activity and performance data; computing an energy dissipation rate based on at least the activity and performance data; and scheduling tasks under the operating system based on the computed energy dissipation rate.
1. A task scheduler, comprising:
a network communication port configured to receive a task;
a respective queue of tasks, for each of a plurality of respective servers configured to process the received task;
an input configured to receive information from each of the plurality of servers relating to a current state dependent on its environment and task processing; and
at least one automated processor configured to:
analyze the queues of tasks, to predict a future state of the plurality of servers, dependent on the current state and as a result of processing the received task having associated resource use characteristics comprising power consumption from the plurality of queues; and
select a respective queue for the received task, dependent on the predicted future state for the plurality of respective servers, a predicted energy efficiency, and a thermal limit criterion.
2. The task scheduler according to claim 1 , wherein the at least one automated processor is further configured to control a cooling system.
3. The task scheduler according to claim 1 , wherein the at least one automated processor is further configured to select a respective queue which concentrates tasks in a queue for a respective server while maintaining operation within a safe operating temperature limit, and meeting at least one performance criterion for the task.
4. The task scheduler according to claim 1 , wherein the at least one automated processor further selects a respective queue for tasks having common associated resource use characteristics, wherein the respective server associated with the queue is selectively configured to efficiently process tasks having the associated resource use characteristics.
5. The task scheduler according to claim 1 , wherein the at least one automated processor employs computational models for predicting a plurality of thermal conditions in a data center to promote overall energy efficiency.
6. The task scheduler according to claim 1 , wherein the at least one automated processor further predictively controls a cooling system for the plurality of servers dependent on tasks in the queue awaiting processing.
7. The task scheduler according to claim 1 , wherein the at least one automated processor is further configured to control an inlet temperature and an air flow pattern of a cooling system.
8. The task scheduler according to claim 1 , wherein the at least one automated processor is further configured to select the respective queue for the received task, further dependent on a latency criterion for the task in the queue.
9. The task scheduler according to claim 1 , wherein the at least one automated processor is further configured to place the task in a respective queue, and to transfer another task previously in the respective queue to a different queue.
10. The task scheduler according to claim 1 , wherein the received information comprises environmental information relating to each respective server.
11. The task scheduler according to claim 1 , wherein the received information comprises current computing load for each respective server.
12. The task scheduler according to claim 1 , wherein the received information comprises a performance metric for each respective server.
13. The task scheduling method according to claim 1 , wherein the respective queue for the received task is selected further dependent on a latency criterion for the task in the queue.
14. The task scheduler according to claim 5 , wherein the computational models are empirical models of energy dissipation dependent on the received information comprising hardware activity counts for the respective servers.
15. A task scheduling method, comprising:
receiving a task through a network communication port;
providing a respective queue of tasks, for each of a plurality of servers configured to process the received task;
receiving information from each of the plurality of servers relating to a current state dependent on its environment and task processing;
analyzing the queues of tasks, to predict a future state for the plurality of servers, dependent on the current state and as a result of processing the received task from the respective queues, each respective received task having associated resource use characteristics comprising power consumption; and
selecting a respective queue for the received task, dependent on predicted energy efficiency, the predicted future state of the plurality of respective servers, and a thermal limit criterion.
16. The task scheduling method according to claim 15 , further comprising proactively controlling a cooling system dependent on unexecuted tasks in the queues.
17. The task scheduling method according to claim 15 , further comprising selecting a respective queue to concentrates tasks for a respective server while maintaining operation within a safe operating temperature limit, and meeting at least one performance criterion for the task.
18. The task scheduling method according to claim 15 , further comprising selecting the respective queue for tasks having common associated resource use characteristics, wherein the respective server associated with the queue is selectively configured to efficiently process tasks having the associated resource use characteristics.
19. The task scheduling method according to claim 16 , wherein the received information comprises hardware activity counts for the respective servers.
20. A task scheduling system, comprising:
a plurality of queues of tasks, each queue of tasks supplying tasks to a respective one of a plurality of respective servers;
an input configured to receive information from each of the plurality of servers relating to a current server state; and
at least one automated processor configured to analyze the plurality of queues of tasks, to predict a future state for each of the plurality of servers as a result of inserting a new task into the queue; and select a respective queue in which to insert the new task dependent on the predicted future state of the plurality of respective servers, a predicted energy efficiency, and a thermal limit criterion.