Neural network model for predicting usage in a hyper-converged infrastructure

1. A method, comprising:

obtaining, by at least one computing device, a training dataset for a neural network executed by the at least one computing device, the training dataset comprising historical data associated with resource consumption of a set of workloads within a hyper-converged infrastructure, wherein the neural network comprises a long short term memory recurrent neural network operating on a time window input, the neural network further utilizing gates comprising a sigmoid neural net layer and a pairwise multiplication operation;

training the neural network based upon the training dataset, the training dataset comprising historical usage of the set of workloads, using a regression analysis applying a least-square error function;

monitoring, by the at least one computing device, usage demand for a particular time period, the usage demand associated with the set of workloads within the hyper-converged infrastructure executed on a plurality of host devices;

providing the usage demand to the neural network;

generating a usage prediction based upon the usage demand, the usage prediction comprising a predicted resource consumption for the set of workloads over a time period;

adjusting an allocation of physical resources within the hyper-converged infrastructure for the set of workloads based upon the usage prediction; and

validating the allocation of the physical resources by performing an ongoing validation of the usage prediction relative to an actual usage demand being placed upon the set of workloads by calculating a similarity measure utilizing a dynamic time warping signal processing algorithm, wherein the neural network can be retrained in response to the similarity measure indicating a difference between the actual usage demand and the usage prediction being greater than a threshold.

2. The method of claim 1 , wherein the resource consumption comprises memory, central processing unit (CPU), and storage usage of the set of workloads.

3. The method of claim 1 , wherein the neural network is trained utilizing a regression model.

4. The method of claim 1 , further comprising:

obtaining, using the at least one computing device, the actual usage demand for the time period.

5. The method of claim 1 , further comprising:

retraining the neural network based upon the similarity between the actual usage demand and the usage prediction.

6. The method of claim 1 , wherein adjusting the allocation of physical resources within the hyper-converged infrastructure comprises increasing or decreasing CPU, memory, or storage resources based upon the usage prediction.

7. The method of claim 1 , wherein the set of workloads comprises one or more workloads, each of the one or more workloads comprising a virtual machine (VM) or an application running on a VM managed by a hypervisor using physical resources of one or more hosts.

8. A system comprising:

at least one computing device;

an application executed by the at least one computing device, the application causing the at least one computing device to at least:

obtain a training dataset for a neural network executed by the at least one computing device, the training dataset comprising historical data associated with resource consumption of a set of workloads within a hyper-converged infrastructure, wherein the neural network comprises a long short term memory recurrent neural network operating on a time window input, the neural network further utilizing gates comprising a sigmoid neural net layer and a pairwise multiplication operation;

train the neural network based upon the training dataset, the training dataset comprising historical usage of the set of workloads, using a regression analysis applying a least-square error function;

monitor usage demand for a particular time period, the usage demand associated with the set of workloads within the hyper-converged infrastructure executed on a plurality of host devices;

provide the usage demand to the neural network;

generate a usage prediction based upon the usage demand, the usage prediction comprising a predicted resource consumption for the set of workloads over a time period;

adjust an allocation of physical resources within the hyper-converged infrastructure for the set of workloads based upon the usage prediction; and

validate the allocation of the physical resources by performing an ongoing validation of the usage prediction relative to an actual usage demand being placed upon the set of workloads by calculating a similarity measure utilizing a dynamic time warping signal processing algorithm, wherein the neural network can be retrained in response to the similarity measure indicating a difference between the actual usage demand and the usage prediction being greater than a threshold.

9. The system of claim 8 , wherein the resource consumption comprises memory, central processing unit (CPU), and storage usage of the set of workloads.

10. The system of claim 8 , wherein the neural network is trained utilizing a regression model.

11. The system of claim 8 , wherein the application causes the at least one computing device to at least:

obtain the actual usage demand for the time period.

12. The system of claim 8 , wherein the application causes the at least one computing device to at least:

retrain the neural network based upon the similarity between the actual usage demand and the usage prediction.

13. The system of claim 8 , wherein the allocation of physical resources within the hyper-converged infrastructure is adjusted by increasing or decreasing CPU, memory, or storage resources based upon the usage prediction.

14. The system of claim 8 , wherein the set of workloads comprises one or more workloads, each of the one or more workloads comprising a virtual machine (VM) or an application running on a VM managed by a hypervisor using physical resources of one or more hosts.

15. A non-transitory computer-readable medium embodying a program executed by at least one computing device, the program causing the at least one computing device to at least:

obtain a training dataset for a neural network executed by the at least one computing device, the training dataset comprising historical data associated with resource consumption of a set of workloads within a hyper-converged infrastructure, wherein the neural network comprises a long short term memory recurrent neural network operating on a time window input, the neural network further utilizing gates comprising a sigmoid neural net layer and a pairwise multiplication operation;

train the neural network based upon the training dataset, the training dataset comprising historical usage of the set of workloads, using a regression analysis applying a least-square error function;

monitor usage demand for a particular time period, the usage demand associated with the set of workloads within the hyper-converged infrastructure executed on a plurality of host devices;

provide the usage demand to the neural network;

generate a usage prediction based upon the usage demand, the usage prediction comprising a predicted resource consumption for the set of workloads over a time period;

adjust an allocation of physical resources within the hyper-converged infrastructure for the set of workloads based upon the usage prediction; and

validate the allocation of the physical resources by performing an ongoing validation of the usage prediction relative to an actual usage demand being placed upon the set of workloads by calculating a similarity measure utilizing a dynamic time warping signal processing algorithm, wherein the neural network can be retrained in response to the similarity measure indicating a difference between the actual usage demand and the usage prediction being greater than a threshold.

16. The non-transitory computer-readable medium of claim 15 , wherein the resource consumption comprises memory, central processing unit (CPU), and storage usage of the set of workloads.

17. The non-transitory computer-readable medium of claim 15 , wherein the neural network is trained utilizing a regression model.

18. The non-transitory computer-readable medium of claim 15 , wherein the program causes the at least one computing device to at least:

obtain the actual usage demand for the time period.

19. The non-transitory computer-readable medium of claim 15 , wherein the program causes the at least one computing device to at least:

retrain the neural network based upon the similarity between the actual usage demand and the usage prediction.

20. The non-transitory computer-readable medium of claim 15 , wherein the set of workloads comprises one or more workloads, each of the one or more workloads comprising a virtual machine (VM) or an application running on a VM managed by a hypervisor using physical resources of one or more hosts.