Scaling a cloud infrastructure

1. A method, comprising:

receiving at least one of resource-level metrics and application-level metrics;

estimating parameters of at least one application based on the at least one of the received resource-level and application-level metrics; and

automatically and dynamically determining directives for scaling application deployment based on the estimated parameters;

wherein at least one of the estimated parameters corresponds to an unobservable parameter of the at least one application and is estimated employing an estimation technique without accessing or modifying the at least one application;

wherein the estimation technique continually adapts to dynamically changing system parameters during runtime; and

wherein the receiving, estimating, and determining steps are performed by a computer system comprising a memory and at least one processor coupled to the memory.

2. The method of claim 1 , wherein the determining is performed in response to changing workload demand.

3. The method of claim 1 , further comprising providing the directives to a cloud service provider to execute the scaling.

4. The method of claim 1 , further comprising determining placement of a virtual machine on a physical machine based on the directives and an availability constraint.

5. The method of claim 1 , wherein the estimation technique comprises a Kalman filtering technique.

6. The method of claim 5 , wherein the Kalman filtering technique comprises specifying a generic queueing-theoretic model.

7. The method of claim 1 , wherein the scaling is performed to meet performance goals.

8. The method of claim 7 , wherein the performance goals are specified in a service level agreement.

9. The method of claim 1 , wherein the determining is performed without user input about dynamically resizing deployment.

10. The method of claim 1 , wherein the directives for scaling comprise one or more directives indicating migration of one or more virtual machines for running the at least one application across one or more physical machines.

11. The method of claim 1 , further comprising determining placement of a virtual machine on a physical machine based on the directives and a colocation constraint.

12. The method of claim 1 , further comprising determining placement of a virtual machine on a physical machine based on the directives and a security constraint.

13. The method of claim 1 , wherein the unobservable parameter comprises a background utilization parameter modeling resource utilization at a server due to jobs running on the server.

14. The method of claim 1 , wherein the unobservable parameter is used to predict future values of server utilization and response time.

15. The method of claim 1 , wherein the unobservable parameter is an existing parameter of the at least one application and hidden from the computer system performing the receiving, estimating, and determining steps.