MECHANISM OF IDENTIFYING AVAILABLE MEMORY RESOURCES IN A NETWORK OF MULTI-LEVEL MEMORY MODULES

• US 20160380921A1
• Filed: 06/23/2015
• Published: 12/29/2016
• Est. Priority Date: 06/23/2015
• Status: Active Grant

• Current Assignee: Advanced Micro Devices, Inc.
• Original Assignee: Advanced Micro Devices, Inc.
• Inventors: Blagodurov, Sergey
• Granted Patent: US 10,171,382 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 3/061 Improving I/O performance
• CPC Class Codes: G06F 3/0631 by allocating resources to storage systems
• CPC Class Codes: G06F 3/067 Distributed or networked storage systems, e.g. storage area networks [SAN], network attached storage [NAS]
• CPC Class Codes: H04L 45/12 Shortest path evaluation
• CPC Class Codes: H04L 47/781 Centralised allocation of resources
• CPC Class Codes: H04L 47/822 Collecting or measuring resource availability data

Abstract

A method of managing memory in a network of nodes includes identifying memory resources for each of the plurality of nodes connected to the network, storing memory resource information describing the memory resources, and based on the stored memory resource information, allocating a portion of the memory resources for execution of instructions in a workload, where at least a first node of the plurality of nodes is configured to execute the workload using the allocated portion of the memory resources.

Claims

1. A method, comprising:

identifying memory resources for each of a plurality of nodes connected to a network;

storing memory resource information describing the memory resources; and

based on the stored memory resource information, allocating a portion of the memory resources for execution of instructions in a workload, wherein at least a first node of the plurality of nodes is configured to execute the workload using the allocated portion of the memory resources.

2. The method of claim 1, wherein identifying the memory resources further comprises, for each node of the plurality of nodes connected to the network, receiving from the node the memory resource information that describes the memory resources of the node.

3. The method of claim 1, wherein the memory resources comprise volatile memory and nonvolatile memory.

4. The method of claim 1, wherein the memory resources comprise one or more of a solid state disk, a hard disk drive, flash memory, phase-change memory, spin-transfer torque memory, and memristor memory.

5. The method of claim 1, wherein for each node of the plurality of nodes, the memory resource information identifies the memory resources controlled by the node, and wherein the method further comprises:

based on the memory resource information, identifying a network link connecting a first memory resource identified by the memory resource information with a second memory resource identified by the memory resource information; and

communicating with the second memory resource via the identified network link.

6. The method of claim 1, further comprising:

in response to determining that a node of the plurality of nodes is unreachable, designating the memory resources controlled by the node as unavailable; and

allocating the portion of the memory resources to replace the unavailable memory resources.

7. The method of claim 1, further comprising, in response to detecting a new node connected to the network, storing memory resource information describing memory resources of the new node.

8. The method of claim 7, further comprising, for each node of the plurality of nodes, determining whether the node is unreachable by periodically polling the node.

9. The method of claim 1, wherein allocating the portion of the memory resources for execution of the first workload further comprises:

identifying the portion of memory resources by performing a search of the stored memory resource information according to one of a plurality of search orders; and

adding the identified portion of the memory resources to a memory hierarchy for the first workload.

10. The method of claim 9, wherein the plurality of search orders includes a depth-first search order and a breadth-first search order, and wherein the method further comprises:

performing the search according to the depth-first search order by traversing a plurality of memory levels controlled by a second node prior to traversing memory levels of any third node; and

performing the search according to the breadth-first search order by traversing a first memory level from each of two or more nodes prior to traversing a second memory level from any of the two or more nodes.

11. The method of claim 9, further comprising, prior to traversing a second node, traversing all nodes reachable from the first node by a communication path including fewer network links than a shortest communication path between the first node and the second node.

12. The method of claim 9, further comprising, in response to determining that a node controlling a subportion of memory resources that is included in the memory hierarchy for the first workload is unreachable by the first node, removing the subportion of memory resources from the memory hierarchy of the first workload, wherein removing the subportion of memory resources comprises, for a virtual memory address associated with the memory hierarchy, changing the destination physical address from a physical memory address corresponding to the subportion of memory resources controlled by the unreachable node to a physical memory address corresponding to a different subportion of memory resources controlled by a node that is capable of communicating with the first node.

13. A system, comprising:

a processor configured to execute a first workload;

a runtime system coupled with the processor, wherein the runtime system is configured to identify memory resources controlled by each of a plurality of nodes connected to a network, and

a memory coupled with the processor, wherein the memory is configured to store memory resource information describing the identified memory resources, wherein the runtime system is further configured to, based on the stored memory resource information, allocate a portion of the identified memory resources for executing instructions in the first workload.

14. The system of claim 13, further comprising the plurality of nodes, wherein each of the plurality of nodes is coupled to at least another node of the plurality of nodes via one or more network links.

15. The system of claim 14, wherein each of the plurality of nodes comprises:

a network interface configured to communicate via one of the at least one or more network links;

a volatile memory coupled with the network adapter; and

a non-volatile memory coupled with the network adapter.

16. The system of claim 15, wherein the non-volatile memory comprises one or more of a solid state disk, a hard disk drive, flash memory, phase-change memory, spin-transfer torque memory, and memristor memory.

17. The system of claim 15, wherein the volatile memory comprises dynamic random access memory (DRAM).

18. The system of claim 14, wherein for each node of the plurality of nodes, the node is configured to:

in response to detecting a connection of the node to the network, provide memory resource information to the runtime system; and

in response to detecting a change in the memory resources controlled by the node, provide updated memory resource information reflecting the change to the runtime system.

19. The system of claim 13, wherein the runtime system is further configured to:

identify the memory resources by receiving the memory resource information, wherein for each node of the plurality of nodes, the memory resource information identifies the node and identifies the memory resources controlled by the node; and

based on the received memory resource information, traverse the network to discover network links connecting the plurality of memory resources.

20. The system of claim 13, further comprising a memory hierarchy for the workload, wherein the memory hierarchy is coupled with the processor, and wherein the runtime is further configured to allocate the portion of the identified memory resources for the execution of the first workload by:

identifying the portion of memory resources by performing a search of the stored memory resource information; and

adding the identified portion of the memory resources to the memory hierarchy.