Machine learning system for workload failover in a converged infrastructure
Systems and methods for analyzing a customer deployment in a converged or hyper-converged infrastructure are disclosed. A machine learning model is trained based upon historical usage data of other customer deployments. A k-means clustering is performed to generate a prediction as to whether a deployment is configured for optimal failover. Recommendations to improve failover performance can also be generated.
1. A method, comprising:
identifying, by at least one computing device, a cluster of virtual machines executed by a plurality of physical host machines in a converged infrastructure environment;
performing, by the at least one computing device, a plurality of simulations for the cluster of virtual machines, the plurality of simulations simulating a failure of one or more physical host machines, the plurality of simulations further simulating an effect on the cluster of virtual machines as a result of the failure of the one or more physical host machines;
generating, by the at least one computing device, a score for each of the simulations, the score representing the effect on the cluster of virtual machines;
performing, by the at least one computing device, a clustering process on one of the simulations based upon on the score and a plurality of policy parameters, the clustering process being trained using data from a plurality of other deployments within the converged infrastructure environment, the clustering process further placing the one of the simulations into one of a plurality of buckets; and
identifying, by the at least one computing device, based on the clustering process, a most similar deployment to the cluster of virtual machines within the converged infrastructure environment.
2. The method of claim 1 , wherein the plurality of buckets comprise one of: a first prediction that the cluster of virtual machines will restart with minimal resource degradation, a second prediction that the cluster of virtual machines will restart with resource degradation, and a third prediction that the one or more of the cluster of virtual machines will not restart.
3. The method of claim 1 , wherein the plurality of policy parameters specify at least one of: a CPU reservation policy for the plurality of physical host machines, a memory reservation policy for the plurality of physical host machines, a failover level policy, and a host-specific failover policy.
4. The method of claim 1 , wherein the clustering process comprises a k-means clustering process performed on the score and the policy parameters.
5. The method of claim 1 , wherein identifying a most similar deployment further comprises:
identifying a configuration of another deployment of a cluster of virtual machines having the smallest Euclidian distance between a first point representing the one of the simulations and a second point representing the other deployment, wherein the other deployment is associated with a prediction that a corresponding cluster of virtual machines associated with the other deployment will restart with minimal resource degradation.
6. The method of claim 5 , further comprising:
generating at least one recommendation to modify the plurality of policy parameters to match a corresponding plurality of policy parameters associated with the other deployment.
7. The method of claim 1 , wherein the score represents an availability and performance score that comprises a representation of a quantity of the virtual machines that will restart and will fail to restart.
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:
identify a cluster of virtual machines executed by a plurality of physical host machines in a converged infrastructure environment;
perform a plurality of simulations for the cluster of virtual machines, the plurality of simulations simulating a failure of one or more physical host machines, the plurality of simulations further simulating an effect on the cluster of virtual machines as a result of the failure of the one or more physical host machines;
generate a score for each of the simulations, the score representing the effect on the cluster of virtual machines; perform a clustering process on one of the simulations based upon on the score and a plurality of policy parameters, the clustering process being trained using data from a plurality of other deployments within the converged infrastructure environment, the clustering process further placing the one of the simulations into one of a plurality of buckets; and
identify based on the clustering process, a most similar deployment to the cluster of virtual machines within the converged infrastructure environment.
9. The system of claim 8 , wherein the plurality of buckets comprise one of: a first prediction that the cluster of virtual machines will restart with minimal resource degradation, a second prediction that the cluster of virtual machines will restart with resource degradation, and a third prediction that the one or more of the cluster of virtual machines will not restart.
10. The system of claim 8 , wherein the plurality of policy parameters specify at least one of: a CPU reservation policy for the plurality of physical host machines, a memory reservation policy for the plurality of physical host machines, a failover level policy, and a host-specific failover policy.
11. The system of claim 8 , wherein the clustering process comprises a k-means clustering process performed on the score and the policy parameters.
12. The system of claim 8 , wherein a most similar deployment is identified by:
identifying a configuration of another deployment of a cluster of virtual machines having the smallest Euclidian distance between a first point representing the one of the simulations and a second point representing the other deployment, wherein the other deployment is associated with a prediction that a corresponding cluster of virtual machines associated with the other deployment will restart with minimal resource degradation.
13. The system of claim 12 , wherein the application further causes the at least one computing device to at least:
generate at least one recommendation to modify the plurality of policy parameters to match a corresponding plurality of policy parameters associated with the other deployment.
14. The system of claim 8 , wherein the score represents an availability and performance score that comprises a representation of a quantity of the virtual machines that will restart and will fail to restart.
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:
identify a cluster of virtual machines executed by a plurality of physical host machines in a converged infrastructure environment;
perform a plurality of simulations for the cluster of virtual machines, the plurality of simulations simulating a failure of one or more physical host machines, the plurality of simulations further simulating an effect on the cluster of virtual machines as a result of the failure of the one or more physical host machines;
generate a score for each of the simulations, the score representing the effect on the cluster of virtual machines;
perform a clustering process on one of the simulations based upon on the score and a plurality of policy parameters, the clustering process being trained using data from a plurality of other deployments within the converged infrastructure environment, the clustering process further placing the one of the simulations into one of a plurality of buckets; and
identify based on the clustering process, a most similar deployment to the cluster of virtual machines within the converged infrastructure environment.
16. The non-transitory computer-readable medium of claim 15 , wherein the plurality of buckets comprise one of: a first prediction that the cluster of virtual machines will restart with minimal resource degradation, a second prediction that the cluster of virtual machines will restart with resource degradation, and a third prediction that the one or more of the cluster of virtual machines will not restart.
17. The non-transitory computer-readable medium of claim 15 , wherein the plurality of policy parameters specify at least one of: a CPU reservation policy for the plurality of physical host machines, a memory reservation policy for the plurality of physical host machines, a failover level policy, and a host-specific failover policy.
18. The non-transitory computer-readable medium of claim 15 , wherein the clustering process comprises a k-means clustering process performed on the score and the policy parameters.
19. The non-transitory computer-readable medium of claim 15 , wherein a most similar deployment is identified by:
identifying a configuration of another deployment of a cluster of virtual machines having the smallest Euclidian distance between a first point representing the one of the simulations and a second point representing the other deployment, wherein the other deployment is associated with a prediction that a corresponding cluster of virtual machines associated with the other deployment will restart with minimal resource degradation.
20. The non-transitory computer-readable medium of claim 19 , wherein the application further causes the at least one computing device to at least:
generate at least one recommendation to modify the plurality of policy parameters to match a corresponding plurality of policy parameters associated with the other deployment.