Cluster resource management in distributed computing systems
AssigneeMicrosoft Technology Licensing, LLC
InventorsManeesh Sah, Rushi Surla, Arild Skjolsvold, Xinhua Ji, Jegan Devaraju
Life of this patent
Estimating the term…
Abstract
Techniques are provided for managing resources among clusters of computing devices in a computing system. Resource reassignment message are generated for indicating that servers are reassigned and in response to resource compute loads exceed or fall below certain thresholds. Techniques also include establishing communications with the reassigned servers to assign compute loads without physically relocating the servers from one cluster to another cluster.
Claims20
A method implemented by a duster resource manager of managing resources among dusters in a distributed computing system, the dusters including first and second clusters individually containing multiple servers interconnected to one another by a computer network and managed by a first cluster controller and a second cluster controller, respectively, the method comprising:
- the cluster resource manager receiving, via the computer network, status data from the first cluster controller of the first cluster, the status data representing a compute load experienced by the first cluster;
- and
- generating and sending a resource removal message to the first cluster controller when the compute load of the first cluster in the status data is below a threshold, the resource removal message indicating that a server from the first cluster is reassigned to the second cluster, the resource removal message instructing the first cluster controller to create or update a configuration file indicating that the server is reassigned from the first cluster, and thereby causing the first cluster controller to subsequently ignore the reassigned server.
A cluster resource manager computing device in a cluster of a distributed computing system having multiple clusters individually containing multiple servers interconnected by a computer network, the computing device comprising:
- a processor;
- and
- a memory containing instructions executable by the processor to cause the processor to:
- receive, via the computer network, status data from a first cluster controller of the first cluster, the status data representing a compute load experienced by the first cluster;
- and
- generate and transmit a resource removal message to the first cluster controller when the compute load of the first cluster in the status data is below a threshold, the resource removal message indicating that a server from the first cluster is reassigned to the second cluster, the resource removal message instructing the first cluster controller to create or update a configuration file indicating that the server is reassigned from the first cluster, and thereby causing the first cluster controller to subsequently ignore the reassigned server.
A method of managing resources among dusters in a distributed computing system, the dusters including first and second dusters, individually containing multiple servers interconnected to one another by a computer network, and managed by a first cluster controller and a second cluster controller, respectively, the method comprising:
- receiving, via the computer network, status data from a first cluster controller of the first cluster, the status data representing a compute load experienced by the first duster is above a threshold;
- and
- generate and transmit a resource reassignment message to the first duster controller at least partially in response to the compute load experienced by the first cluster being above the threshold, the resource reassignment message indicating that a server is being assigned from the second duster to the first duster and for causing a configuration file corresponding to the first cluster to be updated for the first duster to indicate that the server is reassigned from the second cluster to the first duster, without physically relocating the server from the second cluster to the first duster, and for enabling communications between the first duster and the server during subsequent re-initiation of the first duster according to the updated configuration file.
Dependencies are inferred from the claim text.