SOFTWARE ARCHITECTURE FOR SERVICE OF COLLECTIVE MEMORY AND METHOD FOR PROVIDING SERVICE OF COLLECTIVE MEMORY USING THE SAME

• US 20120159115A1
• Filed: 12/08/2011
• Published: 06/21/2012
• Est. Priority Date: 12/17/2010
• Status: Active Grant

• Current Assignee: Electronics and Telecommunications Research Institute
• Original Assignee: Electronics and Telecommunications Research Institute
• Inventors: KIM, Young Ho, JUNG, Sung In, KANG, Dong Jae, CHA, Gyu Il, LIM, Eun Ji
• Granted Patent: US 8,943,294 B2
• Time in Patent Office: Days
• US Class Current: 711/203
• CPC Class Codes: G06F 12/0653 with centralised address assignment

Abstract

Disclosed is a software architecture supporting a large-capacity collective memory layer in a multi-node system by using a remote direct memory access technique and a software virtualization technique and a computing system performing computing processing by using the architecture. In particular, provided is a software architecture including: a memory region managing module collectively managing a predetermined memory region of a node, a memory service providing module providing a large-capacity collective memory service to a virtual address space in a user process, and a memory sharing support module supporting sharing of the large-capacity collective memory of the multi-node system.

Claims

1. A software architecture for a service of a collective volume memory, comprising:

a collective memory managing module virtualizing memories given from nodes to manage the virtualized memories as a collective memory; and

a collective memory service providing module providing the collective memory service by mapping the collective memory to a selected node through a virtual address space.

2. The software architecture of claim 1, further comprising:

a first collective memory service supporting module supporting the collective memory service by providing an interface for using the collective memory;

or

a second collective memory service supporting module supporting the collective memory service by providing a command for processing the collective memory.

3. The software architecture of claim 2, wherein the collective memory service providing module prefetches data in the node to which the collective memory is mapped.

4. The software architecture of claim 1, wherein the collective memory managing module includes:

a contribution memory processing module registering the memories given from the nodes as a contribution memory through a predetermined protocol;

a collective memory generating module virtualizing the registered contribution memory to generate the virtualized contribution memory as the collective memory; and

a virtual engine driving module initializing a virtual engine and managing the generated collective memory with the initialized virtual engine.

5. The software architecture of claim 4, wherein the collective memory managing module further includes:

a service usage condition tracking module tracking a service usage condition of the collective memory;

a first memory page exchanging module exchanging a memory page by using a transaction protocol layer at the time of providing the collective memory service when the collective memory is positioned on a local node according to a tracking result; and

a second memory page exchanging module exchanging the memory page by using an HCA at the time of providing the collective memory service when the collective memory is positioned on a remote node according to the tracking result.

6. The software architecture of claim 5, wherein the collective memory managing module further includes:

a collective memory state information acquiring module acquiring state information associated with the service usage condition of the collective memory through monitoring; and

a collective memory state information reporting module reporting the acquired state information to a manager server managing the collective memory service.

7. The software architecture of claim 1, wherein the collective memory service providing module includes:

an interface processing module processing an interface for supporting the collective memory service;

a collective memory mapping module mapping the collective memory to the address space to be used as the memory for each node selected through the interface; and

a collective memory sharing module allowing the selected nodes to share the collective memory when the number of the selected nodes is at least two.

8. The software architecture of claim 7, wherein the collective memory service providing module further includes:

a page fault generating module generating a page fault at the time when a service using node that intends to use the collective memory service firstly attempts to access a page included in a mapped memory region;

a page fault handling module handling the page fault when the page fault is generated; and

a page cache loading module loading a physical page of the collective memory on a page cache on the service using node.

9. The software architecture of claim 1, wherein the collective memory service providing module reserves and maps the collective memory at a physical memory position of the node giving the memory.

10. The software architecture of claim 1, wherein the software architecture is implemented by the virtual engine shaped through interworking of the nodes or implemented in a kernel of a computing device configuring cloud computing.

11. A method for providing a service of a collective volume memory, comprising:

virtualizing memories given from nodes to manage the virtualized memories as a collective memory; and

providing the collective memory service by mapping the collective memory to a selected node through a virtual address space.

12. The method of claim 11, further comprising:

supporting the collective memory service by providing an interface for using the collective memory or supporting the collective memory service by providing a command for processing the collective memory.

13. The method of claim 12, wherein in the providing of the collective memory service, data is prefetched in the node to which the collective memory is mapped.

14. The method of claim 11, wherein the managing of the collective memory includes:

registering the memories given from the nodes as contribution memories through a predetermined protocol;

virtualizing the registered contribution memories to generate the virtualized contribution memories as a collective memory; and

initializing a virtual engine and managing the generated collective memory with the initialized virtual engine.

15. The software architecture of claim 14, wherein the driving of the virtual engine includes:

tracking a service usage condition of the collective memory; and

exchanging a memory page by using a transaction protocol layer at the time of providing the collective memory service when the collective memory is positioned on a local node according to a tracking result and exchanging the memory page by using an HCA at the time of providing the collective memory service when the collective memory is positioned on a remote node according to the tracking result.

16. The method of claim 15, wherein the managing of the collective memory further includes:

acquiring state information associated with the service usage condition of the collective memory through monitoring; and

reporting the acquired state information to a manager server managing the collective memory service.

17. The method of claim 11, wherein the providing of the collective memory service includes:

processing an interface for supporting the collective memory service;

mapping the collective memory to the address space to be used as the memory for each node selected through the interface; and

allowing the selected nodes to share the collective memory when the number of the selected nodes is at least two.

18. The method of claim 11, wherein the providing of the collective memory service includes:

generating a page fault at the time when a service using node that intends to use the collective memory service firstly attempts to access a page included in a mapped memory region;

handling the page fault when the page fault is generated;

loading a physical page of the collective memory on a page cache on the service using node; and

executing the collective memory service by permitting the service using node to access the collective memory when the loading is completed.