Entropy-based (self-organizing) stability management

• US 8,954,562 B2
• Filed: 09/28/2007
• Issued: 02/10/2015
• Est. Priority Date: 09/28/2007
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Vicente, John
• Primary Examiner(s): Dharia, Rupal
• Assistant Examiner(s): Ma, Wing
• Application Number:
  US11/864,797
• Publication Number:
  US 20090089410A1
• Time in Patent Office: 2,692 Days
• Field of Search: 709/223, 709/224, 370/254
• US Class Current: 709/224
• CPC Class Codes: G06F 9/45533 Hypervisors; Virtual machine monitors
• CPC Class Codes: H04L 41/046 comprising network management agents or mobile agents therefor
• CPC Class Codes: H04L 41/06 Management of faults, events, alarms or notifications
• CPC Class Codes: H04L 41/142 using statistical or mathematical methods
• CPC Class Codes: H04L 41/40 using virtualisation of network functions or resources, e.g. SDN or NFV entities

Abstract

In some embodiments, the invention involves calculating entropy-based stability values to be used in a framework to build a new class of network control (policy) and (state) management services. The framework may be used to build a number of self-management services to support decentralized (mesh) networks. Other embodiments are described and claimed.

Claims

1. A system for entropy-based stability management, comprising:

a hardware processor in a hierarchical network executing a stability agent;

a plurality of computing nodes in a network, each node comprising;

logic comprising an operational agent for autonomous self-promotion and demotion in the network, wherein a promoted node rises to a cluster level above a previous cluster level and becomes part of a virtual cluster at the level above, while remaining in a virtual cluster of the previous level, and wherein a demoted node falls to a cluster level below the previous cluster level and no longer remains in the virtual cluster at the higher level;

logic to provide management and control service functions to cluster k peer nodes in a node that has been promoted to a cluster-head of the cluster k;

logic to provide management and control service functions to other nodes in cluster levels below a current cluster level of the node from which the node has been promoted,

wherein the operational agent for autonomous self-promotion and demotion in the network uses information based on a node'"'"'s behavior relating to (a) reachability, (b) stability and (c) performance efficiency, and wherein the stability behavior is generated by the stability agent,

the stability agent to collect stability information from the plurality of nodes on the network and to determine a relative dependability between and among the plurality of nodes, and to calculate a level of stability of each of the plurality of nodes on the network, wherein the level of stability is calculated using entropy-based statistical methods; and

the operational agent running on at least one node on the network, the operational agent to receive the stability information from the stability agent and to provide management and control operations to at least one aspect of the network, based on the stability information;

wherein the cluster-head of cluster k has no direct visibility of any cluster in a layer directly above cluster k and the cluster-head of cluster k, but no peer node in cluster k, can directly participate in direct inter-cluster communications.

2. The system as recited by claim 1, wherein Shannon'"'"'s Laws of statistical entropy equations are to be applied to each node in a cluster of nodes on the network, in part, using an entropy of node m during the specific period of entropy measurement Δ

T in the cluster C_ij, normalized between [0, 1], to result in a desired measure of stability with respect to node m, in terms of an absolute relative dependability during the period of stability state evaluation, and where measurement of the cluster'"'"'s overall stability, γ

, is defined and captured by a cluster-head associated to C_ij as;

γ

=[H_CH(t,Δ

T)]*min [H_i(t,Δ

T)],

where CH is the cluster-head for C_ij and i =(all nodes in Cij\CH), wherein H_CH (t,Δ

T) represents entropy of node m during the period of measurement;

where node m is cluster-head CH, and where function “

min”

is a minimum entropy of all other nodes in the cluster C_ij.

3. The system as recited in claim 1, wherein the operational agent is to provide management and control of a self-managing network of virtual clusters and where the stability information is to be used with reachability and performance efficiency factors of nodes in the network to enable a determination to be made for promotion and demotion of nodes in the network.

4. The system as recited in claim 1 wherein the management and control operation comprises one or more of routing control, security management and fault management.

22. The system as recited in claim 1, wherein the plurality of computing nodes further comprises logic to rank the reachability of other nodes in the hierarchical network.

23. The system as recited in claim 1, wherein the logic for autonomous self-promotion and demotion in the network is configured to weight the node'"'"'s behavior relating to each of (a) reachability, (b) stability and (c) performance efficiency to rank the node for promotion or demotion, and where the weights are pre-determined based on a function of the network.

25. The system of claim 1, wherein the stability agent is to (a) determine an overall stability for cluster k, and (b) initiate the formation of another cluster based on the determined overall stability for cluster k.

26. The system of claim 1, wherein the promoted node has no direct knowledge of a cluster above the promoted node'"'"'s present cluster level.

5. A system for entropy-based stability management, comprising:

a hardware processor in a hierarchical network executing a stability agent;

a plurality of computing nodes in a network, each node comprising;

logic comprising an operational agent for autonomous self-promotion and demotion in the network, wherein a promoted node rises to a cluster level above a previous cluster level and becomes part of a virtual cluster at the level above, while remaining in a virtual cluster of the previous level, and wherein a demoted node falls to a cluster level below the previous cluster level and no longer remains in the virtual cluster at the higher level;

logic to provide management and control service functions to cluster k peer nodes in a node that has been promoted to a cluster-head of the cluster k;

logic to provide management and control service functions to other nodes in cluster levels below a current cluster level of the node from which the node has been promoted,

wherein the operational agent for autonomous self-promotion and demotion in the network uses information based on a node'"'"'s behavior relating to (a) reachability, (b) stability and (c) performance efficiency, and wherein the stability behavior is generated by the stability agent,

the stability agent to collect stability information from the plurality of nodes on the network and to determine a relative dependability between and among the plurality of nodes, and to calculate a level of stability of each of the plurality of nodes on the network, wherein the level of stability is calculated using entropy-based statistical methods; and

the operational agent running on at least one node on the network, the operational agent to receive the stability information from the stability agent and to provide management and control operations to at least one aspect of the network, based on the stability information;

wherein the cluster-head of cluster k has no direct visibility of any cluster in a layer directly above cluster k and the stability agent executes in one of an embedded partition, virtual machine monitor, virtual appliance, and virtual machine in the processor.

6. The system as recited in claim 5, wherein a reachability agent providing reachability behavior information executes in one or more of an embedded partition, virtual machine monitor, virtual appliance, virtual machine in at least one of the computing nodes on the network, and a performance efficiency agent providing performance efficiency behavior information executes in one or more of an embedded partition, virtual machine monitor, virtual appliance, virtual machine in at least one of the computing nodes on the network, and wherein the reachability agent, performance efficiency agent and stability agent run on either a same or different node, or combination thereof, on the network.

27. The system of claim 5, wherein the cluster-head of cluster k, but no peer node in cluster k, can directly participate in direct inter-cluster communications.

7. A method for entropy-based stability management in a hierarchical network having a plurality of nodes, comprising:

autonomously determining, by a node, whether the node should be autonomously promoted, autonomously demoted or stay at a same level in the network hierarchy, the promotion and demotion of the node is to a current cluster level from a previous cluster level, wherein a promoted node rises to a cluster level above the previous cluster level and becomes part of a virtual cluster at the level above, while remaining in a virtual cluster of the previous level, and wherein a demoted node falls to a cluster level below the previous cluster level and no longer remains in the virtual cluster at the higher level,

wherein information used to enable self-promotion and demotion of the node from the previous cluster is based on collected states, and the information is distributed among nodes in the hierarchical network residing with nodes having a relationship to the previous cluster and wherein the information is not guaranteed to be held by anyone node in the hierarchical network, and wherein membership of the node in a cluster is irrelevant to a physical network model except when related to the node'"'"'s behavior regarding reachability, stability and performance efficiency, and

wherein there is no prohibition or rule against a cluster having only one member or having as many clusters as there are nodes in the network, and wherein cluster membership is based on self-promotion and demotion;

collecting stability information from a plurality of nodes on a hierarchical network, by a stability agent;

determining a relative dependability between and among the plurality of nodes on the network based on the collected stability information, by the stability agent;

calculating, by the stability agent, a level of stability of each of the plurality of nodes on the network, wherein the level of stability is calculated using entropy-based statistical methods; and

providing, by the stability agent, the level of stability information to an operational agent for management and control operations, the operational agent running on a node on the network;

wherein a cluster-head of cluster k has no direct visibility of any cluster in a layer directly above cluster k and the stability agent executes in one of an embedded partition, virtual machine monitor, virtual appliance, and virtual machine in the processor.

8. The method as recited by claim 7, wherein Shannon'"'"'s Laws of statistical entropy equations are to be applied to each node in a cluster of nodes on the network, in part, using an entropy of node m during the specific period of entropy measurement Δ

T in the cluster C_ij, normalized between [0,1], to result in a desired measure of stability with respect to node m, in terms of an absolute relative dependability during the period of stability state evaluation, and where measurement of the cluster'"'"'s overall average stability, γ

, is defined and captured by a cluster-head associated to C_ij as;

γ

=[H_CH(t,Δ

T)]*min [H_i(t,Δ

T)], where CH is the cluster-head for C_ij and I =(all nodes in C_ij\CH), wherein H_CH (t, Δ

T) represents entropy of node m during the period of measurement, where node m is cluster-head CH, and where function “

min”

finds a minimum entropy of all other nodes in the cluster C_ij.

9. The method as recited in claim 7, further comprising:

providing management and control of a self-managing network of virtual clusters, by the operational agent, where the stability information is to be used with reachability and performance efficiency factors of nodes in the network to enable a determination to be made for promotion and demotion of nodes in the network.

10. The method as recited in claim 7, wherein the management and control operations comprises routing control.

11. The method as recited in claim 7, further comprising:

determining by the operational agent executing on a node on the network whether the node is to be promoted, demoted or stay at a same level, wherein a promoted node rises to a cluster level above a previous cluster level and becomes part of a virtual cluster at the level above, while remaining in a virtual cluster of the previous level, and wherein a demoted node falls to a cluster level below the previous cluster level and no longer remains in the virtual cluster at the higher level; and

performing one of a promotion, demotion, or static operation based on results of the determining.

12. The method as recited in claim 11, further comprising:

providing management and control service functions to cluster k peer nodes in a node that has been promoted to the cluster-head of the cluster k;

providing management and control service functions to other nodes in cluster levels below a current cluster level of the node from which the node has been promoted,

wherein the operational agent uses information based on a node'"'"'s behavior relating to (a) reachability, (b) stability and (c) performance efficiency, and wherein the stability behavior is provided in the providing of the level of stability information.

13. The method as recited in claim 7, wherein the stability agent executes in one of an embedded partition, virtual machine monitor, virtual appliance, or virtual machine in a node on the network.

14. The method as recited in claim 13, further comprising:

providing reachability behavior information, by a reachability agent executing in one or more of an embedded partition, virtual machine monitor, virtual appliance, or virtual machine in at least one of the nodes on the network, and

providing performance efficiency behavior information by a performance efficiency agent executing in one or more of an embedded partition, virtual machine monitor, virtual appliance, or virtual machine in at least one of the nodes on the network,

wherein the reachability agent, performance efficiency agent and stability agent run on either a same or different node, or combination thereof, on the network.

15. A non-transitory machine readable storage medium having instructions for entropy-based stability management in a hierarchical network having a plurality of nodes, the instructions stored therein, that when the instructions are executed on a machine, cause the machine to:

autonomously determine, by a node, whether the node should be promoted, demoted or stay at a same level in the network hierarchy, the promotion and demotion of the node is to a current cluster level from a previous cluster level, wherein a promoted node rises to a cluster level above a the previous cluster level and becomes part of a virtual cluster at the level above, while remaining in a virtual cluster of the previous level, and wherein a demoted node falls to a cluster level below the previous cluster level and no longer remains in the virtual cluster at the higher level,

wherein information used to enable self-promotion and demotion of the node from the previous cluster is based on collected states, and the information is distributed among nodes in the hierarchical network residing with nodes having a relationship to the previous cluster and wherein the information is not guaranteed to be held by anyone node in the hierarchical network, and wherein membership of the node in a cluster is irrelevant to a physical network model except when related to the node'"'"'s behavior regarding reachability, stability and performance efficiency, and

wherein there is no prohibition or rule against a cluster having only one member or having as many clusters as there are nodes in the network, and wherein cluster membership is based on self-promotion and demotion;

collect stability information from a plurality of nodes on a network, by a stability agent;

determine a relative dependability between and among the plurality of nodes on the network based on the collected stability information, by the stability agent;

calculate, by the stability agent, a level of stability of each of the plurality of nodes on the network, wherein the level of stability is calculated using entropy-based statistical methods; and

provide, by the stability agent, the level of stability information to an operational agent for management and control operations, the operational agent running on a node on the network

wherein a cluster-head of cluster k has no direct visibility of any cluster in a layer directly above cluster k and the cluster-head of cluster k, but no peer node in cluster k, can directly participate in direct inter-cluster communications.

16. The medium as recited in claim 15, wherein the entropy-based statistical methods are derived from Shannon'"'"'s Laws of statistical entropy equations.

17. The medium as recited by claim 16, wherein the Shannon'"'"'s Laws of statistical entropy equations are to be applied to each node in a cluster of nodes on the network, in part, using an entropy of node m during the specific period of entropy measurement Δ

T in the cluster C_ij, normalized between [0,1], to result in a desired measure of stability with respect to node m, in terms of an absolute relative dependability during the period of stability state evaluation, and where measurement of the cluster'"'"'s overall stability, γ

, is defined and captured by a cluster-head associated to C_ij as;

γ

=[H_CH(t,Δ

T)]*min[H_i(t,Δ

T)], where CH is the cluster-head for C_ij and I =(all nodes in C_ij\CH), wherein H_CH (t, Δ

T) represents entropy of node m during the period of measurement, where node m is cluster-head CH and where function “

min”

finds a minimum entropy of all other nodes in the cluster C_ij.

18. The medium as recited in claim 15, further comprising instructions to:

provide management and control of a self-managing network of virtual clusters, by the operational agent, where the stability information is to be used with reachability and performance efficiency factors of nodes in the network to enable a determination to be made for promotion and demotion of nodes in the network.

19. The medium as recited in claim 15, further comprising instructions to:

determine by the operational agent executing on a node on the network whether the node is to be promoted, demoted or stay at a same level, wherein a promoted node rises to a cluster level above a previous cluster level and becomes part of a virtual cluster at the level above, while remaining in a virtual cluster of the previous level, and wherein a demoted node falls to a cluster level below the previous cluster level and no longer remains in the virtual cluster at the higher level; and

perform one of a promotion, demotion, or static operation based on results of the determining.

20. The medium as recited in claim 19, further comprising instructions to:

provide management and control service functions to cluster k peer nodes in a node that has been promoted to the cluster-head of the cluster k;

provide management and control service functions to other nodes in cluster levels below a current cluster level of the node from which the node has been promoted,

wherein the operational agent uses information based on a node'"'"'s behavior relating to (a) reachability, (b) stability and (c) performance efficiency, and wherein the stability behavior is provided in the providing of the level of stability information.

21. The medium as recited in claim 15, wherein the stability agent executes in one of an embedded partition, virtual machine monitor, virtual appliance, or virtual machine in a node on the network.

24. The medium as recited in claim 15, wherein autonomously determining whether the node should be promoted, demoted or stay at the same level further compromises instructions to weight the node'"'"'s behavior relating to each of (a) reachability, (b) stability and (c) performance efficiency to rank the node for promotion or demotion, and where the weights are pre-determined based on a function of the network.