Automatic demand-driven resource scaling for relational database-as-a-service

• US 10,410,155 B2
• Filed: 05/01/2015
• Issued: 09/10/2019
• Est. Priority Date: 05/01/2015
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Das, Sudipto, Li, Feng, Narasayya, Vivek, Konig, Arnd Christian
• Primary Examiner(s): Brockington, III, William S
• Application Number:
  US14/702,644
• Publication Number:
  US 20160321588A1
• Time in Patent Office: 1,593 Days
• CPC Class Codes: G06F 11/3442 for planning or managing the needed capacity
• CPC Class Codes: G06F 11/3452 Performance evaluation by statistical analysis
• CPC Class Codes: G06F 16/217 Database tuning G06F16/2282 takes precedence; database performance monitoring G06F11/3409
• CPC Class Codes: G06Q 10/06315 Needs-based resource requirements planning or analysis
• CPC Class Codes: G06Q 20/145 Payments according to the detected use or quantity

Abstract

Architecture that enables a Database-as-a-Service (DaaS) to auto-scale container sizes on behalf of tenants. An abstraction is provided that enables tenants to reason about monetary budget and query latency, rather than resource provisioning. An auto-scaling module automatically determines a container size for a subsequent billing interval based on telemetry that comprises latencies (e.g., waits), resource utilizations, and available budget, for example. A set of robust signals are derived from database engine telemetry and combined to significantly improve accuracy of resource demand estimation for database workloads. In a more specific implementation, resource demands can be estimated for arbitrary SQL (structured query language) workloads in a relational database management system (RDBMS).

Claims

1. A database-as-a-service system, comprising:

an auto-scaling module configured to;

automatically compute container sizes of containers in database servers for subsequent billing intervals based on telemetry from the database servers, wherein the container sizes are based on a fixed capacity of resources within a container, wherein the telemetry comprises an available budget for a billing interval, resource utilization, and, for each request for a resource in the container, a time spent waiting for the resource as a percentage of total waits for all resources in the container for an interval of time; and

automatically scale the container sizes of the containers for a subsequent billing interval on behalf of a tenant based on the telemetry; and

at least one hardware processor configured to execute computer-executable instructions in a memory, the instructions executed to enable the auto-scaling module.

2. The system of claim 1, further comprising a telemetry manager configured to collect the telemetry of tenant databases of the database servers.

3. The system of claim 1, wherein the telemetry comprises at least one of latencies or available budget.

4. The system of claim 1, further comprising a resource demand estimator configured to estimate resource demands that warrant a larger container or a smaller container.

5. The system of claim 4, wherein the resource demand estimator is configured to transform the telemetry into signals used to estimate the resource demands of multiple workloads of the database servers.

6. The system of claim 5, wherein the resource demand estimator is configured as a rules-based logic having one or more thresholds for each of the signals to determine tenant state in terms of each signal.

7. The system of claim 1, further comprising a budget manager configured to compute the available budget for each billing interval, which is smaller than a budgeting period, based on budget input criteria.

8. The system of claim 1, further comprising an auto-scaling logic configured to output container sizing actions to a database management fabric based on a combination of raw and derived signals, available budget, and demand estimates.

9. A database-as-a-service method comprising:

estimating resource demands of one or more containers for databases based on production telemetry, wherein the production telemetry comprises an available budget for a billing interval, resource utilization, and, for each request for a resource in the one or more containers, a time spent waiting for the resource as a percentage of total waits for all resources in the one or more containers for an interval of time; and

automatically scaling container sizes of the one or more containers for a subsequent billing interval on behalf of a tenant based on changes in the estimated resource demands, wherein the container sizes are based on a fixed capacity of resources within a container.

10. The method of claim 9, further comprising governing behavior of the automatic scaling based on inputs specified by the tenant.

11. The method of claim 9, further comprising identifying statistically-robust signals from the production telemetry.

12. The method of claim 9, further comprising combining multiple signals to improve accuracy of the estimated resource demands.

13. The method of claim 9, further comprising controlling monetary costs based on monetary budget, performance sensitivity, and latency goals.

14. The method of claim 9, further comprising selecting a subset of counters of the production telemetry from which signals can be derived to estimate resource demands.

15. The method of claim 9, further comprising tracking correlations between signals and trend metrics of counters in the production telemetry.

16. A database-as-a-service method, comprising acts of:

collecting telemetry of database servers, wherein the database servers host containers for corresponding tenant databases, the containers each having corresponding container sizes, wherein the telemetry comprises an available budget for a billing interval, resource utilization, and, for each request for a resource in the containers, a time spent waiting for the resource as a percentage of total waits for all resources in the containers for an interval of time;

estimating resource demands of the containers based on processing of the telemetry; and

automatically scaling the container sizes for a subsequent billing interval in correlation with changes in the estimated resource demands, wherein the container sizes are based on a fixed capacity of resources within a container.

17. The method of claim 16, further comprising governing behavior of the automatic scaling based on inputs specified by a tenant.

18. The method of claim 16, further comprising combining multiple signals to improve accuracy of the estimated resource demands.

19. The method of claim 16, further comprising tracking correlations between signals and trend metrics of counters in the telemetry.

20. The method of claim 16, further comprising deriving signals for estimating the resources demands of the container sizes based on a selection of counters from the telemetry.