High performance computing as a service

Patent metadata

Field	Value
Publication number	US-10089147-B2
Application number	US-85621610-A
Country	US
Kind code	B2
Filing date	Aug 13, 2010
Priority date	Aug 13, 2010
Publication date	Oct 2, 2018
Grant date	Oct 2, 2018

Abstract

A scheduling system and method for high-performance computing (HPC) applications includes a network management component stored in physical memory and executed by a processor. The management component is configured to transform HPC resources into a schedulable resource catalog by transforming multi-dimensional HPC resources into a one dimension versus time resource catalog with a dependent graph structure between resources such that HPC resources are enabled to be provisioned into a service environment with predictable provisioning using the resource catalog. A graphical user interface component is coupled to the network management component and configured to provide scheduling visibility to entities and to enable a plurality of different communication modes for scheduling and communication between entities.

First claim

What is claimed is:

1. A scheduling system for high-performance computing (HPC) applications, comprising:
a network management component stored in physical memory and executed by a processor, the network management component being configured to:
optimize HPC resource utilization by transforming multi-dimensional HPC resources into a schedulable one dimension versus time resource catalog that employs a dependency graph structure stored in physical memory including nodes representing respective ones of the multi-dimensional HPC resources and links between corresponding pairs of nodes representing dependencies between the pairs of nodes such that the multi-dimensional HPC resources are provisioned into a service environment with predictable provisioning using the resource catalog, wherein the multi-dimensional HPC resources include three-dimensional locations and interconnects;
generate and support dynamically-created collaboration and negotiation channels for exchange of information and resources among users; and
filter catalog entries of the resource catalog based on social topology information providing associations among the users,
wherein the filtered catalog entries indicate available multi-dimensional HPC resources to be included in resource negotiations and collaborations, and
wherein the resource catalog is generated by partitioning the multi-dimensional HPC resources and capturing partition dependencies, the partitioning providing optimal speed, performance, and availability of the multi-dimensional HPC resources during the provisioning;

a graphical user interface component coupled to the network management component and configured to provide scheduling visibility to the users, the graphical user interface enabling a plurality of different communication modes for scheduling and communication between the users using a computer network; and
a resource management component configured to manage provisioning of the multi-dimensional HPC resources into the service environment with optimal partition composition using the resource catalog.

2. The system as recited in claim 1 , wherein the resource management component includes a resource wrapper configured to support the plurality of different communication modes and to implement one or more Application Programming Interfaces (APIs) for resource provisioning services, wherein the provisioning services are configured to pre-load application data before a scheduled reservation time to eliminate data access delays for a user at the scheduled reservation time.

3. The system as recited in claim 1 , wherein the plurality of different communication modes includes an interactive mode to enable a user to at least access resource availability and schedules.

4. The system as recited in claim 1 , wherein the plurality of different communication modes includes a batch mode to enable a user to negotiate and reschedule with one or more other users.

5. The system as recited in claim 1 , wherein the resource management component is coupled to the network management component, and wherein the resource management component includes a job packing mechanism configured to job pack the multi-dimensional HPC resources into the resource catalog.

6. The system as recited in claim 5 , wherein the resource management component generates the resource catalog by partitioning the multi-dimensional HPC resources and capturing partition dependencies.

7. The system as recited in claim 1 , further comprising a price component configured to at least balance demand against pricing to populate a schedule.

8. The system as recited in claim 1 , further comprising a communication channel configured to enable the communication between the users.

9. A computer system comprising one or more memories, one or more central processing units, and one or more network connections, and further comprising:
a catalog data structure configured to catalog a plurality of available high performance computing (HPC) resource types into a schedulable one dimension versus time resource catalog that employs a dependency graph structure stored in the one or memories that includes nodes representing respective ones of the HPC resource types and links between corresponding pairs of nodes representing dependencies between the pairs of nodes, wherein the HPC resource types include multi-dimensional HPC resources including three-dimensional locations and interconnects;
a network management component configured to:
generate and support dynamically-created collaboration and negotiation channels for exchange of information and resources among users; and
filter catalog entries of the resource catalog based on social topology information providing associations among the users,
wherein the filtered catalog entries indicate available HPC resource types to be included in resource negotiations and collaborations, and
wherein the resource catalog is generated by partitioning the HPC resource types and capturing partition dependencies, the partitioning providing optimal speed, performance, and availability of the HPC resource types during the provisioning;

a network interface configured to permit one or more requests for the HPC resource types, each of the requests being associated with one or more operations;
a resource inventory mechanism associated with one or more of the nodes, each of the requests being associated with one of the HPC resource types such that when a request for a respective HPC resource type is received, the resource inventory mechanism associated with that request operates on the node associated with the requested HPC resource type to resolve all dependencies defined by the dependency graph structure related to the requested HPC resource type and the operation;
a graphical user interface component coupled to the network management component and configured to provide scheduling visibility to the users, the graphical user interface enabling a plurality of different communication modes for scheduling and communication between the users using a computer network; and
a resource management component configured to manage provisioning of the HPC resource types into a service environment with optimal partition composition based on the resource catalog.

10. The system as recited in claim 9 , wherein the resource inventory mechanism resolves the dependencies by changing respective resource types of nodes to satisfy the request for the respective resource.

11. The system as recited in claim 9 , wherein the HPC resource types include one or more resource attributes.

12. The system as recited in claim 11 , further comprising a pricing process that dynamically tracks one or more prices of the resource attributes.

13. The system as recited in claim 9 , wherein the operations include at least one of: reservation, negotiation, cancellation and view.

14. The system as recited in claim 9 , wherein the social topology information includes information associated with a social network connecting the users.

15. The system as recited in claim 9 , wherein the HPC resource types are associated with the entities by: ownership, control, present use, future allocation, and reservation.

16. The system as recited in claim 9 , further comprising a communication process to enable two or more of the users to reallocate one or more of the associations among the users.

17. The system as recited in claim 9 , wherein at least one of the associations among the users is a dependency defined by the dependency graph structure.

18. A method for scheduling high-performance computing (HPC) applicat

Assignees

• Jamjoom Hani T
• Podlaseck Mark E
• Qu Huiming
• Ruan Yaoping
• Saure Denis R
• Shae Zon Yin
• Sheopuri Anshul
• IBM

Inventors

• Jamjoom Hani T
• Podlaseck Mark E
• Qu Huiming
• Ruan Yaoping
• Saure Denis R
• Shae Zon-Yin
• Sheopuri Anshul

Classifications

• G06F9/5011 Primary

  the resources being hardware resources other than CPUs, Servers and Terminals · CPC title