Method, apparatus and computer program product for scheduling dedicated processing resources

1. A method of scheduling dedicated processing resources, comprising:

in response to receiving a scheduling request for a plurality of dedicated processing resources, obtaining a topology of the plurality of dedicated processing resources, the topology being determined based on connection attributes related to connections among the plurality of dedicated processing resources;

determining, based on the topology, a target dedicated processing resource satisfying the scheduling request from the plurality of dedicated processing resources; and

scheduling the target dedicated processing resource satisfying the scheduling request;

wherein the plurality of dedicated processing resources are distributed across a plurality of dedicated processing resource servers, and obtaining the topology comprises:

obtaining a first connection attribute relating to multiple ones of the dedicated processing resource servers, the first connection attribute indicating at least one of delays, bandwidths, throughputs, transmission rates, transmission qualities and network utilization rates of connections among the plurality of dedicated processing resource servers;

obtaining a second connection attribute relating to a particular one of the dedicated processing resource servers, the second connection attributes indicating types of connections among dedicated processing resources and relevant hardware resources in the particular dedicated processing resource server; and

determining the topology based at least in part on the first and second connection attributes.

2. The method according to claim 1 , wherein determining the topology comprises:

determining the plurality of dedicated processing resource servers as a plurality of nodes in the topology; and

determining, based on the first connection attribute, a distance between two nodes of the plurality of nodes that are connected, the distance indicating a performance of a connection between the two nodes.

3. The method according to claim 2 , wherein determining the topology further comprises:

determining the dedicated processing resources and the relevant hardware resources in the particular dedicated processing resource server as a plurality of sub-nodes of a respective node in the plurality of nodes; and

determining, based on the second connection attribute, a distance between two sub-nodes of the plurality of sub-nodes that are connected, the distance indicating a performance of a connection between the two sub-nodes.

4. The method according to claim 1 , wherein determining the topology comprises:

determining the plurality of dedicated processing resources as a plurality of nodes in the topology; and

determining, based on the first and second connection attributes, a distance between two nodes of the plurality of nodes that are connected, the distance indicating a performance of a connection between the two nodes.

5. The method according to claim 1 , wherein obtaining the topology comprises:

obtaining a utilization rate of each of the plurality of dedicated processing resources; and

determining the topology based on the utilization rates and the connection attributes.

6. The method according to claim 1 , wherein determining the target dedicated processing resource comprises:

determining a group of available dedicated processing resources from the plurality of dedicated processing resources;

obtaining a utilization rate of each dedicated processing resource in the group of available dedicated processing resources;

selecting, based on the utilization rate, a set of dedicated processing resource candidates from the group of available dedicated processing resources; and

determining, based on a required resource amount in the scheduling request, the target dedicated processing resource from the set of dedicated processing resource candidates.

7. The method according to claim 6 , wherein selecting the set of dedicated processing resource candidates comprises:

selecting, from the group of available dedicated processing resources, available dedicated processing resources having utilization rates below a predetermined threshold, as the set of dedicated processing resource candidates.

8. The method according to claim 6 , wherein determining, based on the required resource amount, the target dedicated processing resource comprises:

selecting, based on the topology, dedicated processing resources of which the connections with the set of dedicated processing resource candidates have performances above a predetermined threshold, until the resource amount of the set of dedicated processing resource candidates and the selected dedicated processing resources satisfies the required resource amount.

9. The method according to claim 1 , wherein determining the target dedicated processing resource comprises:

determining a first group of dedicated processing resources and a second group of dedicated processing resources being different from the first group of dedicated processing resources; and

determining the target dedicated processing resource from the first and second groups of dedicated processing resources based on at least one of global load balance, a connection cost, and cross-rack traffic.

10. An apparatus for scheduling dedicated processing resources, comprising:

at least one processing unit;

at least one memory coupled to the at least one processing unit and storing instructions to be executed by the at least one processing unit, the instructions, when executed by the at least one processing unit, causing the apparatus to perform acts comprising:

in response to receiving a scheduling request for a plurality of dedicated processing resources, obtaining a topology of the plurality of dedicated processing resources, the topology being determined based on connection attributes related to connections among the plurality of dedicated processing resources;

determining, based on the topology, a target dedicated processing resource satisfying the scheduling request from the plurality of dedicated processing resources; and

scheduling the target dedicated processing resource satisfying the scheduling request;

wherein the plurality of dedicated processing resources are distributed across a plurality of dedicated processing resource servers, and obtaining the topology comprises:

obtaining a first connection attribute relating to multiple ones of the dedicated processing resource servers, the first connection attribute indicating at least one of delays, bandwidths, throughputs, transmission rates, transmission qualities and network utilization rates of connections among the plurality of dedicated processing resource servers;

obtaining a second connection attribute relating to a particular one of dedicated processing resource servers, the second connection attribute indicating types of connections among dedicated processing resources and relevant hardware resources in the particular dedicated processing resource server; and

determining the topology based at least in part on the first and second connection attributes.

11. The apparatus according to claim 10 , wherein determining the topology comprises:

determining the plurality of dedicated processing resource servers as a plurality of nodes in the topology; and

determining, based on the first connection attribute, a distance between two nodes of the plurality of nodes that are connected, the distance indicating a performance of a connection between the two nodes.

12. The apparatus according to claim 11 , wherein determining the topology further comprises:

determining the dedicated processing resources and the relevant hardware resources in the particular dedicated processing resource server as a plurality of sub-nodes of a respective node in the plurality of nodes; and

determining, based on the second connection attribute, a distance between two sub-nodes of the plurality of sub-nodes that are connected, the distance indicating a performance of a connection between the two sub-nodes.

13. The apparatus according to claim 10 , wherein determining the topology comprises:

determining the plurality of dedicated processing resources as a plurality of nodes in the topology; and

determining, based on the first and second connection attributes, a distance between two nodes of the plurality of nodes that are connected, the distance indicating a performance of a connection between the two nodes.

14. The apparatus according to claim 10 , wherein obtaining the topology comprises:

obtaining a utilization rate of each of the plurality of dedicated processing resources; and

determining the topology based on the utilization rate and the connection attributes.

15. The apparatus according to claim 10 , wherein determining the target dedicated processing resource comprises:

determining a group of available dedicated processing resources from the plurality of dedicated processing resources;

obtaining a utilization rate of each dedicated processing resource in the group of available dedicated processing resources;

selecting, based on the utilization rate, a set of dedicated processing resource candidates from the group of available dedicated processing resources; and

determining, based on a required resource amount in the scheduling request, the target dedicated processing resource from the set of dedicated processing resource candidates.

16. The apparatus according to claim 15 , wherein selecting the set of dedicated processing resource candidates comprises:

selecting, from the group of available dedicated processing resources, available dedicated processing resources having utilization rates below a predetermined threshold, as the set of dedicated processing resource candidates.

17. The apparatus according to claim 15 , wherein determining, based on the required resource amount; the target dedicated processing resource comprises:

selecting, based on the topology, dedicated processing resources of which the connections with the set of dedicated processing resource candidates have performance above a predetermined threshold, until the resource amount of the set of dedicated processing resource candidates and the selected dedicated processing resources satisfies the required resource amount.

18. A computer program product comprising a non-transitory computer readable storage medium having stored therein machine executable instructions of one or more software programs, wherein the machine executable instructions, when executed by a machine, cause the machine to:

in response to receiving a scheduling request for a plurality of dedicated processing resources, obtain a topology of the plurality of dedicated processing resources, the topology being determined based on connection attributes related to connections among the plurality of dedicated processing resources;

determine, based on the topology, a target dedicated processing resource satisfying the scheduling request from the plurality of dedicated processing resources; and

schedule the target dedicated processing resource satisfying the scheduling request;

wherein the plurality of dedicated processing resources are distributed across a plurality of dedicated processing resource servers, and obtaining the topology comprises:

obtaining a first connection attribute relating to multiple ones of the dedicated processing resource servers, the first connection attribute indicating at least one of delays, bandwidths, throughputs, transmission rates, transmission qualities and network utilization rates of connections among the plurality of dedicated processing resource servers;

obtaining a second connection attribute relating to a particular one of dedicated processing resource servers, the second connection attribute indicating types of connections among dedicated processing resources and relevant hardware resources in the particular dedicated processing resource server; and

determining the topology based at least in part on the first and second connection attributes.

19. The computer program product according to claim 18 , wherein determining the topology comprises:

determining the plurality of dedicated processing resource servers as a plurality of nodes in the topology; and

determining, based on the first connection attribute, a distance between two nodes of the plurality of nodes that are connected, the distance indicating a performance of a connection between the two nodes.

20. The computer program product according to claim 19 , wherein determining the topology further comprises:

determining the dedicated processing resources and the relevant hardware resources in the particular dedicated processing resource server as a plurality of sub-nodes of a respective node in the plurality of nodes; and

determining, based on the second connection attribute, a distance between two sub-nodes of the plurality of sub-nodes that are connected, the distance indicating a performance of a connection between the two sub-nodes.