MECHANISM FOR EFFICIENT DISCOVERY OF STORAGE RESOURCES IN A RACK SCALE ARCHITECTURE SYSTEM

• US 20180034909A1
• Filed: 07/28/2016
• Published: 02/01/2018
• Est. Priority Date: 07/28/2016
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Nachimuthu, Murugasamy K., Kumar, Mohan J.
• Granted Patent: US 10,791,174 B2
• Time in Patent Office: Days
• CPC Class Codes: H04L 41/0813 characterised by the conditions triggering a change of settings
• CPC Class Codes: H04L 41/0895 Configuration of virtualised networks or elements, e.g. virtualised network function or OpenFlow elements
• CPC Class Codes: H04L 67/1097 for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]

Abstract

Mechanisms for efficient discovery of storage resources in a Rack Scale Architecture (RSA) system and associated methods, apparatus, and systems. A rack is populated with pooled system drawers including pooled compute drawers and pooled storage drawers communicatively coupled via input-output (IO) cables. Compute nodes including one or more processors, memory resources, and optional local storage resources are installed in the pooled compute drawers, and are enabled to be selectively-coupled to storage resources in the pooled storage drawers over virtual attachment links. During a discovery process, a compute node determines storage resource characteristics of storage resources it may be selectively-coupled to and the attachment links used to access the storage resources. The storage resource characteristics are aggregated by a pod manager that uses corresponding configuration information to dynamically compose compute nodes for rack users based on user needs.

Claims

1. A computer-implemented method, comprising:

for each of a plurality of compute nodes in a pooled compute drawer installed in a rack,

discovering storage resource characteristics for one or more storage resources in a pooled storage drawer installed in the rack and coupled in communication with the compute node via a respective attachment link between the compute node and the one or more storage resources;

communicating the storage resource characteristics that are discovered to a configuration management entity for the rack; and

determining, via the configuration management entity for the rack, the storage resource characteristics for each compute node.

2. The method of claim 1, wherein the storage resources in the pooled storage drawer include a plurality of storage drives, and wherein the storage resource characteristics include storage drive health information for at least a portion of the plurality of storage drives.

3. The method of claim 1, wherein the storage resource characteristics include a storage capacity for each of the one or more storage resources and attachment link performance characteristics for each attachment link.

4. The method of claim 1, further comprising:

for a compute node in the pooled compute drawer, performing a storage discovery process including,

initializing an attachment link over which a storage resource in the pooled storage drawer is accessible to the compute node;

acquiring capacity and health information for the storage resource accessible to the compute node over the attachment link; and

determining storage attachment link characteristics for the attachment link.

5. The method of claim 4, wherein the pooled compute drawer in which the compute node is installed includes a drawer manager, the method further comprising:

performing the storage discovery process for each of multiple compute nodes in the first pooled compute drawer;

providing to the drawer manager, for each of the multiple compute nodes, capacity and health information for the one or more storage resources that are accessible to that compute node via one or more respective attachment links used by the compute node to access storage resources in the first pooled storage drawer and the storage attachment link characteristics for each attachment link.

6. The method of claim 5, further comprising:

in response to a discovery event or discovery request,

requesting, via the drawer manager, the compute node to perform the storage discovery process.

7. The method of claim 6, wherein the discovery event is one of:

a) a compute node being reset;

b) a new compute node being installed in the first pooled compute drawer; and

c) a compute node being recomposed.

8. The method of claim 4, wherein a compute node includes one or more local storage resources, and the method further includes determining local storage resource characteristics for the compute node that include at least one of performance and health characteristics for each of the one or more local storage resources and communicating the local storage resource characteristics for the compute node to the configuration management entity.

9. The method of claim 1, wherein each of the plurality of compute nodes includes at least one processor and has access to at least one of local memory and shared memory from a shared memory pool, the method further comprising:

for each compute node;

communicating performance characteristics for each of the at least one processor to the configuration management entity; and

communicating capacity and performance characteristics for the at least one of local memory and shared memory accessed by the compute node to the configuration management entity.

10. The method of claim 1, wherein an attachment link comprises a virtual link including one or more Peripheral Component Interconnect Express (PCIe) link segments traversing at least one PCIe switch.

11. The method of claim 1, wherein the rack includes multiple pooled compute drawers, each having a configuration similar to the first pooled compute drawer, and multiple pooled storage drawers, each having a configuration similar to the first pooled storage drawer, and wherein the method further comprises:

for each of a plurality of compute nodes in each of the pooled compute drawers,

automatically discovering storage resource characteristics for one or more storage resources in a pooled storage drawer coupled in communication with the compute node via a respective attachment link between the compute node and the one or more storage resources;

communicating the storage resource characteristics that are discovered to a configuration management entity for the rack; and

aggregating, via the configuration management entity for the rack, the storage resource characteristics for each compute node.

12. A pooled compute drawer, configured to be installed in a rack including a pooled storage drawer having a plurality of storage resources coupled to an input-output (IO) switch and including at least one IO port, the pooled compute drawer comprising:

a plurality of compute nodes, each compute node including,

one or more processors;

one or more memory resources, operatively coupled to the processor; and

an IO interface;

one or more IO ports, each IO port configured to be coupled to an end of a respective IO cable that has an opposing end coupled to a respective IO port in the pooled storage drawer when the pooled compute drawer is installed in the rack; and

IO infrastructure including a IO interconnect, configured to couple each compute node to at least one of the one or more IO ports;

wherein each of the plurality of compute nodes is configured to access one or more storage resource in the pooled storage drawer when the pooled compute drawer is installed in the rack, the access to a storage resource for a given compute node being facilitated over a virtual attachment link coupling the compute node in communication with the storage resource,

and wherein each of the plurality of compute nodes is further configured to discover storage resource characteristics for one or more storage resources in the pooled storage drawer the compute node is enabled to access via one or more respective attachment links coupling the compute node to the one or more storage resource, wherein the storage resource characteristics include a storage capacity for each of the one or more storage resources and attachment link performance characteristics for each of the one or more attachment links.

13. The pooled compute drawer of claim 12, wherein the storage resources in the first pooled storage drawer include a plurality of storage drives, and wherein the storage resource characteristics include at least one of storage drive health information and storage drive performance information.

14. The pooled compute drawer of claim 12, wherein a compute node is configured to perform a storage discover process comprising:

for each of one or more storage resources in the pooled storage drawer,

initializing an attachment link coupling the compute node in communication with the storage resource;

acquiring capacity and health information for the storage resource; and

determining attachment link characteristics for the attachment link.

15. The pooled compute drawer of claim 12, further comprising a drawer manager, wherein each compute node is configured to provide storage resource characteristics including capacity and health information for the one or more storage resources that are accessible to that compute node via the one or more attachment links used by the compute node to access storage resources in the pooled storage drawer and the storage attachment link characteristics for each of the one or more attachment links to the drawer manager.

16. The pooled compute drawer of claim 15, wherein the drawer manager is configured to request that a compute node perform a discovery process in response to one or more of the following events:

a) a compute node being reset,

b) a new compute node being installed in the pooled compute drawer, and

c) a compute node being recomposed.

17. The pooled compute drawer of claim 15, wherein each compute node is configured to:

communicate performance characteristics for each of its one or more processors to the drawer manager; and

communicate capacity and performance characteristics for its one or more memory resources to the drawer manager.

18. The pooled compute drawer of claim 15, wherein one or more of the compute nodes includes at least one local storage resource, and each of the one or more compute nodes is configured to report at least one of capacity and health information for the at least one local storage resource included in that compute node.

19. The pooled compute drawer of claim 15, wherein the rack includes a pod manager that is communicatively coupled to the drawer manager when the pooled compute drawer is installed in the rack, and wherein the drawer manager is further configured to forward, for each compute node, the storage resource characteristics provided to the drawer manager from that compute node to the pod manager.

20. A rack comprising:

a rack frame, having a plurality of pooled system drawers installed therein, including,

a pooled compute drawer including,

a plurality of compute nodes, each having one or more processors and one or more memory resources,

at least one input-output (IO) port; and

IO infrastructure communicatively coupled each compute node to an IO port;

a pooled storage drawing including,

a plurality of storage resources;

at least one IO port; and

IO infrastructure communicatively coupling each storage resource to at least one IO port; and

at least one IO cable coupled between respective IO ports on the pooled compute drawer and the pooled storage drawer,

wherein each compute node is configured to be selectively-coupled to one or more storage resources in the pooled storage drawer over a respective attachment link coupling the compute node in communication with a respective storage resource,

and wherein each of the plurality of compute nodes is further configured to discover storage resource characteristics for the one or more storage resources in the pooled storage drawer the compute node is configured to be selectively-coupled to, wherein the storage resource characteristics for each storage resource includes a storage capacity for the storage resource and attachment link performance characteristics for the attachment link over which the compute node is configured to be selectively-coupled.

21. The rack of claim 20, wherein the storage resources in the pooled storage drawer include a plurality of storage drives, and wherein the storage resource characteristics include at least one of storage drive health information and storage drive performance information.

22. The rack of claim 20, wherein a compute node is configured to perform a storage discovery process comprising:

initializing an attachment link coupling the compute node in communication with a storage resource in the pooled storage drawer;

acquiring capacity and health information for the storage resource; and

determining storage attachment link characteristics for the attachment link.

23. The rack of claim 20, wherein the rack includes a pod manager and the pooled compute drawer includes a drawer manager communicatively coupled to the pod manager, wherein each compute node is configured to provide to the drawer manager the storage resource characteristics including capacity and health information for the one or more storage resources in the pooled storage drawer to which the compute node is configured to be selectively-coupled, and wherein the drawer manager is configured to send the storage resource characteristics it is provided by each compute node to the pod manager.

24. The rack of claim 23, wherein the drawer manager is configured to request that a compute node perform a discovery process in response to one or more of the following events:

a) a compute node being reset,

b) a new compute node being installed in the pooled compute drawer, and

c) a compute node being recomposed for a different client.

25. The rack of claim 23, wherein each compute node is configured to:

communicate performance characteristics for each of its one or more processors to the drawer manager; and

communicate capacity and performance characteristics for its one or more memory resources to the drawer manager.

26. At least one non-transitory machine readable medium having instructions stored thereon, configured to be executed in a distributed manner on a plurality of processing entities in a rack in which a plurality pooled system drawers are installed, wherein distributed execution of the instructions on the plurality of processing entities performs operations comprising:

for each of a plurality of compute nodes in a pooled compute drawer installed in a rack,

discovering storage resource characteristics for one or more storage resources in a pooled storage drawer installed in the rack and coupled in communication with the compute node via a respective attachment link between the compute node and the one or more storage resources;

communicating the storage resource characteristics that are discovered to a configuration management entity for the rack; and

determining, via the configuration management entity for the rack, the storage resource characteristics for each compute node.

27. The least one non-transitory machine readable medium of claim 26, wherein the storage resources in the pooled storage drawer include a plurality of storage drives, and wherein the storage resource characteristics include storage drive health information for at least a portion of the plurality of storage drives.

28. The least one non-transitory machine readable medium of claim 26, wherein the storage resource characteristics include a storage capacity for each of the one or more storage resources and attachment link performance characteristics for each attachment link.

29. The least one non-transitory machine readable medium of claim 26, wherein distributed execution of the instructions performs further operations, comprising:

for a compute node in the pooled compute drawer, performing a storage discovery process including,

initializing an attachment link over which a storage resource in the pooled storage drawer is accessible to the compute node;

acquiring capacity and health information for the storage resource accessible to the compute node over the attachment link; and

determining storage attachment link characteristics for the attachment link.

30. The least one non-transitory machine readable medium of claim 29, wherein the first pooled compute drawer in which the compute node is installed includes a drawer manager, and wherein distributed execution of the instructions performs further operations, comprising:

performing the storage discovery process for each of multiple compute nodes in the first pooled compute drawer;

providing to the drawer manager, for each of the multiple compute nodes, capacity and health information for the one or more storage resources that are accessible to that compute node via one or more respective attachment links used by the compute node to access storage resources in the first pooled storage drawer and the storage attachment link characteristics for each attachment link.