DYNAMIC ALLOCATION OF RESOURCES WHILE CONSIDERING RESOURCE RESERVATIONS

• US 20140379924A1
• Filed: 05/06/2014
• Published: 12/25/2014
• Est. Priority Date: 06/21/2013
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Das, Sudipto, Narasayya, Vivek, Syamala, Manoj, Li, Feng
• Granted Patent: US 9,602,426 B2
• Time in Patent Office: Days
• US Class Current: 709/226
• CPC Class Codes: H04L 47/72 using reservation actions during connection setup
• CPC Class Codes: H04L 67/10 in which an application is distributed across nodes in the network software deployment G06F8/60; multiprogramming arrangements G06F9/46

Abstract

Described herein are technologies relating to computing resource allocation among multiple tenants. Each tenant may have a respective absolute reservation for rate-based computing resources, which is independent of computing resource reservations of other tenants. The multiple tenants vie for the rate-based computing resources, and tasks are scheduled based upon which tenants submit the tasks and the resource reservations of such tenants.

Claims

1. A method comprising:

receiving, at a computing device that has a fixed amount of rate-based computing resources, a indications of a plurality of tenants that are configured to request rate-based resources of a computing device, a tenant being logically isolated computer-executable code;

receiving, at the computing device, respective resource reservations for the plurality of tenants, the resource reservations being absolute and cumulatively greater than the fixed amount of computing resources of the computing device, a resource reservation of the respective resource reservations defining an amount of rate-based computing resources that is to be accessible to a tenant associated with the resource reservation when the tenant requests rate-based computing resources;

receiving requests for rate-based computing resources of the computing device from tenants in the plurality of tenants; and

dynamically allocating rate-based computing resources to the tenants based upon the requests and the respective resource reservations for the plurality of tenants.

2. The method of claim 1, a first tenant in the plurality of tenants has a first resource reservation, the first resource reservation being a static reservation, the static reservation indicating that the first tenant has reserved a first threshold amount of rate-based computing resources of the computing device for each metering interval.

3. The method of claim 2, the static reservation indicating that the first tenant is not guaranteed access to rate-based computing resources of the computing device above the first threshold amount of rate-based computing resources at any time.

4. The method of claim 2, a second tenant in the plurality of tenants has a second resource reservation, the second resource reservation being an elastic reservation, the elastic reservation indicating that the second tenant has reserved a second threshold amount of rate-based computing resources of the computing device over several metering intervals.

5. The method of claim 4, the elastic reservation indicating that the second tenant is to have access to rate-based computing resources of the computing device above the second threshold amount of rate-based computing resources when the second tenant has not utilized the second threshold amount of rate-based computing resources over the several metering intervals.

6. The method of claim 4, the first threshold amount of rate-based computing resources and the second threshold amount of rate-based computing resources are expressed as one of respective percentages of the fixed amount of rate-based computing resources, respective numbers of virtual cores, or respective CPU times.

7. The method of claim 1, further comprising:

computing respective fractional deficits for the requests; and

allocating the rate-based resources based upon the fractional deficits computed for the requests.

8. The method of claim 1, further comprising:

receiving an indication of at least one of respective prices paid by owners of the plurality of tenants for the respective resource reservations or penalties respectively corresponding to the plurality of tenants; and

dynamically allocating the rate-based computing resources to the tenants based upon the at least one of the prices or the penalties.

9. The method of claim 8, wherein the penalties define respective reimbursements to be made to the plurality of tenants when the respective resource reservations are not met.

10. The method of claim 8, further comprising:

setting a value that is indicative of a fairness criteria to employ when allocating the rate-based computing resources to the tenants when the rate-based computing resources requested by the tenants are in accord with the resource reservations of the tenants and when the rate-based computing resources requested by the tenants exceed the fixed amount of rate-based computing resources of the computing device; and

dynamically allocating the rate-based computing resources to the tenants based upon the value.

11. The method of claim 1, further comprising reimbursing an owner of at least one tenant in the tenants when a resource reservation of the tenant is unmet.

12. A server computing device, comprising:

a processor core; and

a memory that comprises a scheduler component that is executed by the processor core, the scheduler component comprises;

a queue manager component that maintains tasks that are executable by the processor core, the tasks respectively generated by tenants of the server computing device, the tenants having respective resource reservations for rate-based computing resources of the server computing device; and

a task selector component that selects a task from the tasks to be executed by the processor core based upon the respective resource reservations of the tenants, the processor core executes the task responsive to the task selector component selecting the task.

13. The server computing device of claim 12, wherein a first resource reservation in the resource reservations is a static resource reservation, and wherein a second resource reservation in the reservations is an elastic resource reservation.

14. The server computing device of claim 12, the scheduler component further comprises a deficit determiner component that receives respective computing resource utilizations of the tenants and the respective resource reservations of the tenants and computes respective fractional deficits for the tenants based upon the computing resource utilizations and the resource reservations, the task selector component selects the task based upon the fractional deficits.

15. The server computing device of claim 14, the respective computing resource utilizations of the tenants being with respect to metering intervals for the tenants, the metering intervals being time intervals during which the computing resource utilizations are computed and accounted.

16. The server computing device of claim 14, the task selector component selects the task based upon a tenant that submitted the task having a largest fractional deficit amongst the fractional deficits.

17. The server computing device of claim 14, the memory further comprises a meter component that computes the respective computing resource utilizations of the tenants over multiple processor cores.

18. The server computing device of claim 12, the task being a database task.

19. The server computing device of claim 12, the task selector component selects the task to at least one of maximize revenue of an operator of the server computing device or to contemplate fairness with respect to tenants with workloads.

20. A computer-readable storage medium comprising instructions that, when executed by a processor, cause the processor to perform acts comprising:

receiving a plurality of tasks that are to be executed by a processor core, the plurality of tasks respectively correspond to a plurality of tenants, each tenant in the plurality of tenants being respective computer-executable code that generates tasks, the plurality of tenants have a respective plurality of resource reservations that include both static and elastic resource reservations;

computing respective resource utilizations of the plurality of tenants over respective metering intervals for the plurality of tenants, the metering intervals being respective time periods over which the respective resource utilizations of the plurality of tenants are computed;

computing respective fractional deficits for the plurality of tenants based upon the respective resource utilizations and the respective resource reservations;

selecting a task from the plurality of tasks based upon the respective fractional deficits; and

responsive to the task being selected, executing the task.