Control loop for scaling agents in a cluster

1. A computer-implemented method, the method comprising:

receiving, at a controller in a feedback loop, a target number of idle agents, wherein an idle agent is a computing resource in a cluster of cloud-based computing resources;

receiving, at the controller, a current number of idle agents in the cluster of cloud-based computing resources;

computing a control variable for the cluster of cloud-based computing resources based on the target number of idle agents and the current number of idle agents; and

submitting the control variable to a scheduler for managing the cluster of cloud-based computing resources, wherein the number of idle agents in the cluster of cloud-based computing resources scales based on the control variable,

wherein the controller is a proportional-integral-derivative controller, and the control variable is a sum of:

a proportional term that is proportional to an error, wherein the error is a difference between the target number of idle agents and the current number of idle agents;

an integral term proportional to a sum of past values of the error; and

a derivative term proportional to a change in the error.

2. The computer-implemented method of claim 1 , wherein the controller is configured to store past values of the error in a moving window, and the integral term is calculated based on the past values in the moving window.

3. The computer-implemented method of claim 2 , wherein the integral term is calculated from an exponentially weighted moving average of values in the moving window.

4. The computer-implemented method of claim 1 , further comprising:

comparing the integral term to a threshold value; and

if the integral term exceeds the threshold value, calculating the control variable as a sum of the proportional term and the threshold value.

5. The computer-implemented method of claim 1 , further comprising:

comparing the control variable to a threshold value; and

if the control variable exceeds the threshold value, submitting the threshold value to the scheduler.

6. The computer-implemented method of claim 1 , wherein the cluster is for performing tasks associated with a vehicle software build.

7. The computer-implemented method of claim 1 , wherein the cluster is for performing continuous integration build tasks.

8. A computer-implemented system, comprising:

one or more processing units; and

one or more non-transitory computer-readable media storing instructions, when executed by the one or more processing units, cause the one or more processing units to perform operations comprising:

receiving a target number of idle agents, wherein an idle agent is a computing resource in a cluster of cloud-based computing resources;

receiving a current number of idle agents in the cluster of cloud-based computing resources;

computing a control variable for the cluster of cloud-based computing resources based on the target number of idle agents and the current number of idle agents; and

submitting the control variable to a scheduler for managing the cluster of cloud-based computing resources, wherein the number of idle agents in the cluster of cloud-based computing resources scales based on the control variable,

wherein the control variable is a sum of:

a proportional term that is proportional to an error, wherein the error is a difference between the target number of idle agents and the current number of idle agents;

an integral term proportional to a sum of past values of the error; and

a derivative term proportional to a change in the error.

9. The computer-implemented system of claim 8 , wherein the operations further comprise storing past values of the error in a moving window, and the integral term is calculated based on the past values in the moving window.

10. The computer-implemented system of claim 8 , the operations further comprising:

comparing the integral term to a threshold value; and

if the integral term exceeds the threshold value, calculating the control variable as a sum of the proportional term and the threshold value.

11. The computer-implemented system of claim 10 , the operations further comprising:

comparing the control variable to a threshold value; and

if the control variable exceeds the threshold value, submitting the threshold value to the scheduler.

12. The computer-implemented system of claim 8 , wherein the cluster is for performing continuous integration build tasks.

13. One or more non-transitory, computer-readable media encoded with instructions that, when executed by one or more processing units, cause the one or more processing units to perform operations comprising:

receiving, at a controller in a feedback loop, a target number of idle agents, wherein an idle agent is a computing resource in a cluster of cloud-based computing resources;

receiving, at the controller, a current number of idle agents in the cluster of cloud-based computing resources;

computing a control variable for the cluster of cloud-based computing resources based on the target number of idle agents and the current number of idle agents; and

submitting the control variable to a scheduler for managing the cluster of cloud-based computing resources, wherein the number of idle agents in the cluster of cloud-based computing resources scales based on the control variable,

wherein the control variable is a sum of:

a proportional term that is proportional to an error, wherein the error is a difference between the target number of idle agents and the current number of idle agents;

an integral term proportional to a sum of past values of the error; and

a derivative term proportional to a change in the error.

14. The one or more non-transitory, computer-readable media of claim 13 , wherein the controller is configured to store past values of the error in a moving window, and the integral term is calculated based on the past values in the moving window.