Methods and systems for migrating network resources to improve network utilization

• US 9,141,437 B2
• Filed: 01/29/2008
• Issued: 09/22/2015
• Est. Priority Date: 01/29/2008
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Krishnamurthy, Rajaram B.
• Primary Examiner(s): Murray, Daniel C
• Assistant Examiner(s): COONEY, ADAM A
• Application Number:
  US12/021,341
• Publication Number:
  US 20090193122A1
• Time in Patent Office: 2,793 Days
• Field of Search: 709/226
• US Class Current: 1/1
• CPC Class Codes: G06F 9/5088 involving task migration
• CPC Class Codes: G06F 9/526 Mutual exclusion algorithms
• CPC Class Codes: H04L 43/0876 Network utilisation, e.g. volume of load or congestion level
• CPC Class Codes: H04L 47/823 Prediction of resource usage
• CPC Class Codes: H04L 47/83 based on usage prediction

Abstract

A method and system are disclosed for migrating network resources to improve network utilization, for use in a multi-node network wherein nodes of the network share network resources. The method comprises the steps of identifying a group of nodes that share one of the network resources, and identifying one of the nodes satisfying a specified condition based on at least one defined access latency metric. The shared resource is moved to the identified one of the nodes to reduce overall access latency to access the shared resource by said group of nodes. One embodiment of the invention provides a method and system to synchronize tasks in a distributed computation using network attached devices (NADs). A second embodiment of the invention provides a method and system to reduce lock latency and network traffic by migrating lock managers to coupling facility locations closest to nodes seeking resource access.

Claims

1. A method of migrating network resources to improve network utilization, for use in a multi-node network wherein nodes of the network share network resources, the method comprising the steps of:

identifying a group of nodes that share one of the network resources;

for each of a plurality of the nodes of the multi-node network, determining an associated aggregate access latency time, based on at least one defined access latency metric, indicating an aggregate latency time for requests from the nodes of said group of nodes, in aggregate, to reach said each node of said plurality of nodes;

identifying one of the nodes from the multi-node network based on said aggregate access latency times;

moving said shared resource to a location in the network based on the identified one of the nodes; and

after moving said shared resource to said location, directing requests from said group of nodes for the shared resource to the identified one of the nodes to reduce overall aggregate access latency time for said requests from said group of nodes to reach the shared resource.

2. A method according to claim 1, wherein the moving step includes the steps of:

moving said shared resource from a home node to said identified one of the nodes; and

said home node forwarding requests for the shared resource to said identified one of the nodes.

3. A method according to claim 2, wherein the step of moving the shared resource from the home node includes the steps of:

said home node keeping track of the number of accesses to said shared resource;

when the shared resource is accessed a given number of times, the home node moving said shared resource to said identified one of the nodes;

providing each of the client nodes with a resource indirection table including information to identify the switch or NAD to contact to request a particular resource; and

providing the home node with a resource mapping table to identify a resource id, the current resource owner, the resource manager location and list of requesting nodes.

4. A method according to claim 2, wherein the step of moving said shared resource from the home node includes the step of using said home node to move the shared resource to said identified one of the nodes.

5. A method according to claim 1, wherein said defined metric is based on sum of node hop counts from the group of nodes that share the resources to other nodes in the network.

6. A method according to claim 1, wherein said resource migration is based on one of (i) being specified explicitly at compile-time by a programmer or software compiler of the system and (ii) during run-time by implicitly detecting spatial and temporal locality in resource access for a single resource or a combination of resources.

7. A method according to claim 1, wherein:

the group of nodes are a first group of computers located in a first datacenter;

said first group of computers access resources owned by a second group of computers located in a second datacenter; and

the shared resource is a lock manager that manages access to the resources of the second group of computers; and

the moving step includes the step of moving the lock manager from the second datacenter to the first datacenter.

8. A method according to claim 7, wherein:

the first group of computers are connected together by a first coupling facility;

the second group of computers are connected together by a second coupling facility; and

the moving step includes the step of moving the lock manager from the second coupling facility to the first coupling facility.

9. A method according to claim 8, wherein the first group of computers access the lock manager on the first coupling facility for access to the resources of the second group of computers in the second datacenter, and the lock manager is moved from the second datacenter to the first datacenter if the only requestors of the lock manager are computers in the first datacenter.

18. The method according to claim 1, wherein:

the determining includes, for each of said plurality of nodes, determining an amount of time, based on said at least one defined latency metric, needed for the requests from each of the nodes of said group of nodes to access reach said each of the nodes of said plurality of nodes, and summing said amounts of time for the requests from all of said nodes of said group of nodes to obtain an associated total access latency time for said each node of said plurality of nodes; and

said identifying includes identifying said one of the nodes based on the associated total access latency times obtained for said plurality of nodes.

19. The method according to claim 18, wherein said identifying includes identifying the one of the nodes having the minimum of the associated total access latency times.

20. The method according to claim 1, wherein:

the aggregate access latency time is based on the distance between each of the group of nodes and each of the plurality of nodes, as measured according to a defined distance metric.

10. A system for migrating network resources to improve network utilization, for use in a multi-node network wherein nodes of the network share network resources, the system comprising:

one or more processing units configured for (i) identifying a group of nodes that share one of the network resources, (ii) for each of a plurality of the nodes of the multi-node network, determining an associated aggregate access latency time, based on at least one defined access latency metric, indicating an aggregate latency time for requests from the nodes of said group of nodes, in aggregate, to reach said each node of said plurality of nodes;

(iii) identifying one of the nodes based on said aggregate access latency times, and (iv) moving said shared resource to the identified one of the nodes, and (v) after moving said shared resource to said location, directing requests from said group of nodes for the shared resource to the identified one of the nodes to reduce overall aggregate access latency time for said requests from said group of nodes to reach the shared resource.

11. A system according to claim 10, wherein:

said moving is done by moving said shared resource from a home node to said identified one of the nodes; and

said one or more processor units are further configured for forwarding requests for the shared resource from the home node to said identified one of the nodes.

12. A system according to claim 11, wherein said one or more processor units are further configured for keeping track of the number of accesses to said shared resource at the home node;

and when the shared resource is accessed a given number of times at the home node, moving said shared resource to said identified one of the nodes.

13. A system according to claim 10, wherein:

the group of nodes are a first group of computers located in a first datacenter;

said first group of computers are connected together by a first coupling facility at the first datacenter;

said first group of computers access a second group of computers located in a second datacenter;

the second group of computers are connected together by a second coupling facility at the second datacenter;

the shared resource is a lock manager that manages access to other resources of the second group of computers; and

one or more processor units are further configured for moving the lock manager from the second coupling facility to the first coupling facility.

14. An article of manufacture comprising a program storage device readable by computer, tangibly embodying a program of instructions executable by the computer to perform method steps for migrating network resources to improve network utilization, for use in a multi-node network wherein nodes of the network share network resources, the method steps comprising:

identifying a group of nodes that share one of the network resources;

for each of a plurality of the nodes of the multi-node network, determining an associated aggregate access latency time, based on at least one defined access latency metric, indicating an aggregate latency time for requests from the nodes of said group of nodes, in aggregate, to reach said each node of said plurality of nodes;

identifying one of the nodes based on said access latency times;

moving said shared resource to the identified one of the nodes; and

after moving said shared resource to said location, directing requests from said group of nodes for the shared resource to the identified one of the nodes to reduce overall access latency time for said requests from said group of nodes to reach the shared resource.

15. An article of manufacture according to claim 14, wherein the moving step includes the steps of:

moving said shared resource from a home node to said identified one of the nodes; and

said home node forwarding requests for the shared resource to said identified one of the nodes.

16. An article of manufacture according to claim 14, wherein said defined metric is based on the latency access to the shared resources by said group of nodes.

17. An article of manufacture according to claim 14, wherein said defined metric is based on temporal locality in resource access for a single resource or a combination of resources.