Systems and methods for energy efficient load balancing at server clusters
AssigneeIntel Corporation
InventorsRen Wang, Sanjay Rungta, Tsung-Yuan Tai, Chih-Fan Hsin, Jr-Shian Tsai
Life of this patent
Estimating the term…
Abstract
Methods and systems to balance the load among a set of processing units, such as servers, in a manner that allows the servers periods of low power consumption. This allows energy efficient operation of the set of processing units. Moreover, the process is adaptable to variations in systemic response times, so that systemic response times may be improved when operational conditions so dictate.
Claims20
A method, comprising:
- dispatching, from a load balancer to a processing unit, consecutive requests for a processing unit service, wherein the number of consecutive requests dispatched to the processing unit represents a batch having a size equal to a batch size, and subsequent batches of requests are grouped by batch size and dispatched to successive respective processing units in round robin fashion one batch size per processing unit, wherein processing units that are not processing a batch enter a low power state;
- monitoring a systemic response time for servicing of the requests;
- and
- when the systemic response time varies from an acceptable range, modifying the batch size for a subsequent dispatching.
A computer program product including a non-transitory computer readable medium having computer program logic stored therein, the computer program logic including:
- logic to cause a processor to dispatch, from a load balancer to a processing unit, consecutive requests for a processing unit service, wherein the number of consecutive requests dispatched to the processing unit represents a batch having a size that equals a batch size, and to further cause the processor to dispatch subsequent batches of requests grouped by batch size to successive respective processing units in round robin fashion one batch size per processing unit, wherein processing units that are not processing a batch enter a low power state;
- logic to cause the processor to monitor a systemic response time for servicing of the requests;
- and
- logic to cause the processor to modify the batch size for subsequent dispatching, when the systemic response time varies from an acceptable range.
A system, comprising:
- a processor;
- and
- a memory in communication with said processor, wherein the memory stores a plurality of processing instructions configured to direct said processor to
- dispatch, from a load balancer to a processing unit, consecutive requests for a processing unit service, wherein the number of consecutive requests dispatched to the processing unit represents a batch having a size equal to a batch size, and subsequent batches of requests are grouped by batch size and are dispatched to successive respective processing units in round robin fashion one batch size per processing unit, wherein processing units that are not processing a batch enter a low power state;
- monitor a systemic response time for servicing of the requests;
- and
- modify the batch size for subsequent dispatching when the systemic response time varies from an acceptable range.
Dependencies are inferred from the claim text.