Applying machine learning to dynamically scale computing resources to satisfy a service level agreement (SLA)

1. A method comprising: receiving, by a service, a job request, the job request comprising:

a job,

a priority of the job, and

a callback that identifies an application to execute the job;

adding, by the service, the job request to a queue;

determining, by a machine learning algorithm comprising ordinary least squares, a risk score associated with the job, the risk score comprising a probability that completing the job will not satisfy a set of pre-defined constraints, the risk score determined based on a plurality of factors comprising:

the priority associated with the job; and

currently queued jobs in the queue including a number of the currently queued jobs and an individual priority associated with individual jobs in the currently queued jobs;

re-ordering the currently queued jobs in the queue based on a priority associated with each of the currently queued jobs in the queue;

based on determining that the job is at a head of the queue, invoking the callback to call the application to execute the job, wherein the application is hosted by one or more virtual machines of a plurality of virtual machines being executed in a cloud hosting facility; and

based on receiving an indication that the job has been completed:

marking the job as completed;

determining an environment state comprising:

the priority of the job;

an initial place of the job in the queue; and

a number of virtual machines used to execute other jobs ahead of the job in the queue; and

retraining the machine learning algorithm based at least in part on the environment state.

2. The method of claim 1 , further comprising:

in response to determining that the risk score is between a first threshold and a second threshold, allocating additional resources;

in response to determining that the risk score is between the second threshold and a third threshold, provisioning additional virtual machines; and

in response to determining that the risk score exceeds the third threshold, sending a notification to an administrator.

3. The method of claim 1 , further comprising:

receiving a second job request comprising a second job and a second application;

provisioning one or more additional virtual machines based at least in part on a second risk score associated with the second application; and

after determining that the second application associated with the second job has completed executing the second job, instructing the cloud hosting facility to de-provision at least one virtual machine of the one or more additional virtual machines.

4. The method of claim 1 , further comprising:

determining, by the service, that a difference between a capacity of the plurality of virtual machines of the cloud hosting facility and an estimated job effort associated with the currently queued jobs in the queue satisfies a system determined threshold; and

sending a de-provisioning request, by the service, to the cloud hosting facility to de-provision at least one virtual machine of the plurality of virtual machines.

5. The method of claim 1 , further comprising:

based on determining that a difference between a capacity of the plurality of virtual machines of the cloud hosting facility and an estimated job effort associated with completing the currently queued jobs in the queue fails to satisfy a system determined threshold, sending a provisioning request to the cloud hosting facility to provision a number of additional virtual machines, the number of additional virtual machines determined at in least in part on the difference.

6. The method of claim 1 , wherein:

a the set of pre-defined constraints are specified by a service level agreement associated with the job.

7. The method of claim 1 , wherein the plurality of factors further comprise:

an estimated number of resources and virtual machines to complete the currently queued jobs;

one or more date-time related factors;

additional criteria provided by the application in the job request; and

a number of currently provisioned virtual machines.

8. A service to perform operations comprising: receiving a job request comprising:

a job,

a priority of the job, and

a callback that identifies an application to execute the job;

adding the job request to a queue;

determining, using a machine learning algorithm comprising ordinary least squares, a risk score associated with the job, the risk score comprising a probability that completing the job will not satisfy a set of pre-defined constraints, the risk score determined based on a plurality of factors comprising:

the priority associated with the job; and

currently queued jobs in the queue including a number of the currently queued jobs and an individual priority associated with individual jobs in the currently queued jobs;

re-ordering the currently queued jobs in the queue based on a priority associated with each of the currently queued jobs in the queue;

based on determining that the job is at a head of the queue, invoking the callback to call the application to execute the job, wherein the application is hosted by one or more virtual machines of a plurality of virtual machines being executed in a cloud hosting facility; and

based on receiving an indication that the job has been completed:

marking the job as completed;

determining an environment state comprising:

the priority of the job;

an initial place of the job in the queue; and

an amount of virtual machines used to execute other jobs ahead of the job in the queue; and

retraining the machine learning algorithm based at least in part on the environment state.

9. The service of claim 8 , the operations further comprising:

in response to determining that the risk score is between a first threshold and a second threshold, allocating additional resources;

in response to determining that the risk score is between the second threshold and a third threshold, provisioning additional virtual machines; and

in response to determining that the risk score exceeds the third threshold, sending a notification to an administrator.

10. The service of claim 8 , the operations further comprising:

receiving a second job request comprising a second job and a second application;

provisioning one or more additional virtual machines based at least in part on a second risk score associated with the second application; and

after determining that the second application associated with the second job has completed executing the second job instructing the cloud hosting facility to de-provision at least one virtual machine of the one or more additional virtual machines.

11. The service of claim 8 , the operations further comprising:

based on determining that a difference between a capacity of the plurality of virtual machines of the cloud hosting facility and an estimated job effort associated with completing the currently queued jobs in the queue satisfies a predetermined threshold instructing the cloud hosting facility to de-provision at least one virtual machine of the plurality of virtual machines.

12. The service of claim 8 , wherein:

a the set of pre-defined constraints are specified by a service level agreement associated with the job, wherein the job request identifies the service level agreement.

13. The service of claim 8 , wherein the plurality of factors further comprise:

an estimated number of resources and virtual machines to complete the currently queued jobs;

one or more date-time related factors;

one or more additional factors supplied within the job request; and

a number of currently provisioned virtual machines.

14. One or more non-transitory computer-readable media storing instructions that are executable by one or more processors to perform operations comprising:

receiving, by a service, a job request, the job request comprising:

a job,

a priority of the job, and

a callback that identifies an application to execute the job;

adding, by the service, the job request to a queue;

determining, by a machine learning algorithm comprising ordinary least squares, a risk score associated with the job, the risk score comprising a probability that completing the job will not satisfy a set of pre-defined constraints, the risk score determined based on a plurality of factors comprising:

the priority associated with the job; and

currently queued jobs in the queue including a number of the currently queued jobs and an individual priority associated with individual jobs in the currently queued jobs;

re-ordering the currently queued jobs in the queue based on a priority associated with each of the currently queued jobs in the queue;

based on determining that the job is at a head of the queue, invoking the callback to call the application to execute the job, wherein the application is hosted by one or more virtual machines of a plurality of virtual machines being executed in a cloud hosting facility; and

based on receiving an indication that the job has been completed:

marking the job as completed;

determining an environment state comprising:

the priority of the job;

an initial place of the job in the queue; and

an amount of virtual machines used to execute other jobs ahead of the job in the queue; and

retraining the machine learning algorithm based at least in part on the environment state.

15. The one or more non-transitory computer-readable media of claim 14 , the operations further comprising:

in response to determining that the risk score is between a first threshold and a second threshold, allocating additional resources;

in response to determining that the risk score is between the second threshold and a third threshold, provisioning additional virtual machines; and

in response to determining that the risk score exceeds the third threshold, sending a notification to an administrator.

16. The one or more non-transitory computer-readable media of claim 14 , the operations further comprising:

receiving a second job request comprising a second job and a second application;

provisioning one or more additional virtual machines based at least in part on a second risk score associated with the second application; and

after determining that the second application has completed executing the second job, instructing the cloud hosting facility to de-provision at least one virtual machine of the one or more additional virtual machines.

17. The one or more non-transitory computer-readable media of claim 14 , the operations further comprising:

determining that a difference between a capacity of the plurality of virtual machines of the cloud hosting facility and an estimated job effort associated with completing the currently queued jobs in the queue of the service satisfies a predetermined threshold; and

sending a de-provisioning request to the cloud hosting facility to de-provision at least one virtual machine of the plurality of virtual machines.

18. The one or more non-transitory computer-readable media of claim 14 , wherein:

the set of pre-defined constraints are specified by a service level agreement associated with the job.

19. The one or more non-transitory computer-readable media of claim 14 , wherein the plurality of factors further comprise:

an estimated number of resources and virtual machines to complete the currently queued jobs;

one or more date-time related factors;

one or more additional factors supplied within the job request; and

a number of currently provisioned virtual machines.

20. The one or more non-transitory computer-readable media of claim 14 , the operations further comprising:

re-ordering the currently queued jobs in the queue based at least in part on:

an individual risk score associated with each queued job of the currently queued jobs; and

an individual priority associated with each queued job of the currently queued jobs.