Intelligent self-configuring decision method for the cloud resources management and orchestration

1. A method for reducing the monitored-data load and the processing load/delay on a cloud management system (CMS), the method comprising:

applying a prediction process to provide predicted key performance indicator (KPI) values for a plurality of VMs managed by the CMS during a first monitoring epoch;

collecting, during the first monitoring epoch, observed KPI values for the plurality of VMs managed by the CMS;

assessing the accuracy of the prediction process by way of calculating, according to a reward function, reward values for the plurality of VMs based on a deviation between the observed KPI values and the predicted KPI values;

calculating a monitoring frequency for collecting monitoring information during a second monitoring epoch based on the reward values; and

collecting the monitoring information during the second monitoring epoch according to the calculated monitoring frequency,

wherein the monitoring information collected during the second monitoring epoch includes second epoch KPI values.

2. The method of claim 1 , further comprising collecting, according to an initial monitoring frequency and an initial monitoring epoch duration set to operator specific default values, initial KPI values for the plurality of VMs.

3. The method of claim 2 , wherein the prediction process is a machine learning method trained by the initial KPI values.

4. The method of claim 1 , wherein the KPIs include network usage, computation usage, and memory usage.

5. The method of claim 1 further comprising calculating a second monitoring epoch duration for collecting monitoring information during the second monitoring epoch based on the reward values or a next decision time,

wherein collecting the monitoring information during the second monitoring epoch according to the calculated monitoring frequency is performed for the second monitoring epoch duration.

6. The method of claim 1 , wherein the reward function depends on a mean square error between the predicted KPI values and the observed KPI values.

7. The method of claim 1 , further comprising providing VM profiles for the plurality of VMs managed by the CMS.

8. The method of claim 7 , wherein the VM profiles are each represented by a vector that specifies a consolidated value for storage resources, computations resources, and network resources required by a respective VM.

9. The method of claim 7 , further comprising clustering the plurality of VM profiles into a plurality of VM affinity groups.

10. The method of claim 9 , wherein clustering the plurality of VM profiles into a plurality of VM affinity groups includes using an enhanced k-means algorithm to solve an optimization problem.

11. The method of claim 10 , further comprising allocating the VMs among a plurality of compute nodes according to the VM affinity groups in order to balance a load on each of the compute nodes.

12. The method of claim 11 , wherein allocating the VMs among a plurality of servers is performed according to an affinity-aided VM scheduling algorithm that takes into account general scheduling information of the VM affinity groups and applies the second epoch KPI values.

13. The method of claim 9 , further comprising increasing the number of affinity groups in response to a sub-threshold number of movements of VMs from one affinity group to another during a period of time.

14. The method of claim 9 , further comprising decreasing the number of affinity groups in response more than a threshold number of movements of VMs from one affinity group to another during a period of time.

15. A cloud management system (CMS) comprising computer readable media including instructions performing a method for reducing the monitored-data load and the processing load/delay, the method comprising:

applying a prediction process to provide predicted key performance indicator (KPI) values for a plurality of VMs managed by the CMS during a first monitoring epoch;

collecting, during the first monitoring epoch, observed KPI values for the plurality of VMs managed by the CMS;

assessing the accuracy of the prediction process by way of calculating, according to a reward function, reward values for the plurality of VMs based on a deviation between the observed KPI values and the predicted KPI values; and

calculating a monitoring frequency for collecting monitoring information during a second monitoring epoch based on the reward values; and

collecting the monitoring information during the second monitoring epoch according to the calculated monitoring frequency,

wherein the monitoring information collected during the second monitoring epoch includes second epoch KPI values.