Apparatus, method, and computer program for coordinating interdependent closed loops
The disclosure relates to an apparatus comprising means for: receiving, at a first closed loop function from a second closed loop function, one or more proposed actions to be taken by the second closed loop function with respect to a managed entity, the first and second closed loop functions operating on the managed entity; determining, at the first closed loop function, one or more further actions to be taken, wherein the determining is based on the at least the one or more proposed actions to be taken by the second closed loop function; and causing, by the first closed loop function, the determined one or more further actions to be taken.
1 . An apparatus for coordinating resource autoscaling in a virtualized network infrastructure, the apparatus comprising:
a processor; and
a memory storing computer-executable instructions that, when executed by the processor, cause the apparatus to operate as an intermediate closed loop (CL) function in a grouped hierarchical closed loop instance having a predefined closedLoopGroupInstanceUniqueId, the grouped hierarchical closed loop instance comprising:
an innermost closed loop configured for vertical virtual network function (VNF) autoscaling of central processing unit (CPU) resources of a VNF,
the intermediate closed loop configured exclusively for horizontal VNF autoscaling of the same VNF, and
an outermost closed loop configured exclusively for node autoscaling of a physical node hosting the VNF, wherein the intermediate closed loop and the innermost closed loop operate on the same managed entity comprising the VNF and its CPU resources; and
wherein the instructions further cause the apparatus to perform the following operations:
receiving, via an Insight Reporting and Control Interface, from a Decision stage of the innermost closed loop, a structured insight report comprising:
(i) an identifier of the managed entity including the VNF instance identifier and CPU scope,
(ii) a predicted vertical scaling action comprising adding or removing a specified number of CPU units,
(iii) an associated execution timestamp, and
(iv) an accuracy or confidence value of the predicted vertical scaling action;
applying a machine reasoning or machine learning algorithm using:
(i) the received structured insight report,
(ii) historical VNF performance data collected in a Collection stage of the intermediate closed loop, and
(iii) configuration data stored in a Knowledge component shared within the grouped hierarchical closed loop instance,
based on the received predicted vertical scaling action and the applied machine reasoning or machine learning algorithm, determining the following:
(i) a horizontal scaling action comprising instantiating or releasing one or more VNF instances, and
(ii) a control action directed to the innermost closed loop comprising pausing or unpausing a Decision stage of the innermost closed loop;
transmitting, via the Insight Reporting and Control Interface, a control message to the innermost closed loop that sets or releases a pause point in the Decision stage of the innermost closed loop when the determined control action includes pausing or unpausing; and
causing execution of the determined horizontal scaling action toward the managed entity via an Execution stage of the intermediate closed loop, wherein, during execution of the determined horizontal scaling action, the Decision stage of the innermost closed loop remains paused to prevent conflicting autoscaling actions within the grouped hierarchical closed loop instance.
2 . The apparatus of claim 1 , wherein the structured insight report is received as a streaming data feed including periodic updates at a predefined time granularity of less than five minutes.
3 . The apparatus of claim 1 , wherein the horizontal scaling action comprises instantiating an additional VNF instance on the same physical node hosting the VNF prior to execution of the predicted vertical scaling action.
4 . The apparatus of claim 1 , wherein the Insight Reporting and Control Interface is further configured to update a closedLoopGoal attribute of the innermost closed loop to modify CPU scaling threshold levels before unpausing the Decision stage.
5 . The apparatus of claim 4 , wherein the machine reasoning or machine learning algorithm comprises a deep neural network trained using historical CPU utilization time-series data of the VNF stored in the Knowledge component.
6 . The apparatus of claim 5 , wherein the control message sets the pause point in the Decision stage of the innermost closed loop for a predefined pause duration corresponding to the execution timestamp of the horizontal scaling action.
7 . The apparatus of claim 6 , wherein the machine reasoning or machine learning algorithm further evaluates the accuracy or confidence value of the predicted vertical scaling action against a predefined confidence threshold and determines the control action when the confidence value is below the predefined confidence threshold.
8 . The apparatus of claim 7 , wherein the Knowledge component shared within the grouped hierarchical closed loop instance stores scaling threshold parameters for CPU utilization, memory usage, and network throughput associated with the VNF.
9 . The apparatus of claim 8 , wherein the intermediate closed loop aggregates multiple structured insight reports from the innermost closed loop over a sliding time window prior to applying the machine reasoning or machine learning algorithm.
10 . The apparatus of claim 9 , wherein the grouped hierarchical closed loop instance is configured such that only one Decision stage among the innermost, intermediate, and outermost closed loops is permitted to execute at a time, and the remaining Decision stages are paused during execution of the horizontal scaling action.
11 . A method for coordinating resource autoscaling in a virtualized network infrastructure by an apparatus configured to operate as an intermediate closed loop (CL) function in a grouped hierarchical closed loop instance having a predefined closedLoopGroupInstanceUniqueId, the grouped hierarchical closed loop instance comprising:
an innermost closed loop configured for vertical virtual network function (VNF) autoscaling of central processing unit (CPU) resources of a VNF,
the intermediate closed loop configured exclusively for horizontal VNF autoscaling of the same VNF, and
an outermost closed loop configured exclusively for node autoscaling of a physical node hosting the VNF, wherein the intermediate closed loop and the innermost closed loop operate on the same managed entity comprising the VNF and its CPU resources,
the method comprising:
receiving, via an Insight Reporting and Control Interface, from a Decision stage of the innermost closed loop, a structured insight report comprising:
(i) an identifier of the managed entity including the VNF instance identifier and CPU scope,
(ii) a predicted vertical scaling action comprising adding or removing a specified number of CPU units,
(iii) an associated execution timestamp, and
(iv) an accuracy or confidence value of the predicted vertical scaling action;
applying a machine reasoning or machine learning algorithm using:
(i) the received structured insight report,
(ii) historical VNF performance data collected in a Collection stage of the intermediate closed loop, and
(iii) configuration data stored in a Knowledge component shared within the grouped hierarchical closed loop instance,
based on the received predicted vertical scaling action and the applied machine reasoning or machine learning algorithm, determining the following:
(i) a horizontal scaling action comprising instantiating or releasing one or more VNF instances, and
(ii) a control action directed to the innermost closed loop comprising pausing or unpausing a Decision stage of the innermost closed loop;
transmitting, via the Insight Reporting and Control Interface, a control message to the innermost closed loop that sets or releases a pause point in the Decision stage of the innermost closed loop when the determined control action includes pausing or unpausing; and
causing execution of the determined horizontal scaling action toward the managed entity via an Execution stage of the intermediate closed loop, wherein, during execution of the determined horizontal scaling action, the Decision stage of the innermost closed loop remains paused to prevent conflicting autoscaling actions within the grouped hierarchical closed loop instance.
12 . The method of claim 11 , wherein the structured insight report is received as a streaming data feed including periodic updates at a predefined time granularity of less than five minutes.
13 . The method of claim 11 , wherein the horizontal scaling action comprises instantiating an additional VNF instance on the same physical node hosting the VNF prior to execution of the predicted vertical scaling action.
14 . The method of claim 11 , wherein the Insight Reporting and Control Interface is further configured to update a closedLoopGoal attribute of the innermost closed loop to modify CPU scaling threshold levels before unpausing the Decision stage.
15 . The method of claim 14 , wherein the machine reasoning or machine learning algorithm comprises a deep neural network trained using historical CPU utilization time-series data of the VNF stored in the Knowledge component.
16 . The method of claim 15 , wherein the control message sets the pause point in the Decision stage of the innermost closed loop for a predefined pause duration corresponding to the execution timestamp of the horizontal scaling action.
17 . The method of claim 16 , wherein the machine reasoning or machine learning algorithm further evaluates the accuracy or confidence value of the predicted vertical scaling action against a predefined confidence threshold and determines the control action when the confidence value is below the predefined confidence threshold.
18 . The method of claim 17 , wherein the Knowledge component shared within the grouped hierarchical closed loop instance stores scaling threshold parameters for CPU utilization, memory usage, and network throughput associated with the VNF.
19 . The method of claim 18 , wherein the intermediate closed loop aggregates multiple structured insight reports from the innermost closed loop over a sliding time window prior to applying the machine reasoning or machine learning algorithm.
20 . The method of claim 19 , wherein the grouped hierarchical closed loop instance is configured such that only one Decision stage among the innermost, intermediate, and outermost closed loops is permitted to execute at a time, and the remaining Decision stages are paused during execution of the horizontal scaling action.