Modeling and forecasting reserve capacity for overbooked clusters

Patent metadata

Field	Value
Publication number	US-10225337-B2
Application number	US-201615199590-A
Country	US
Kind code	B2
Filing date	Jun 30, 2016
Priority date	Jun 30, 2016
Publication date	Mar 5, 2019
Grant date	Mar 5, 2019

Abstract

Managing cluster resources for a cluster. The method includes identifying a plurality of quiescence policies. Each of the quiescence policies define one or more conditions for cluster entities which when met, identify that cluster resources should be reclaimed from cluster entities by quiescing cluster entities. The method further includes, for each quiescence policy, identifying an amount of reclaimed capacity defining the amount of cluster resources that will be reclaimed by implementing the quiescence policy. The method further includes, for each quiescence policy, identifying an amount of reserve capacity defining the amount of cluster resources that should be reserved for resuming quiesced cluster entities. The method further includes; selecting a quiescence policy based on comparing the reclaimed capacities and reserve capacities for the quiescence policies. The method further includes, booking cluster resources for cluster entities on the cluster according to the selected quiescence policy.

First claim

What is claimed is:

1. A computer system comprising:
one or more processors; and
one or more computer-readable media having stored thereon instructions that are executable by the one or more processors to configure the computer system to perform a method of managing cluster resources for a cluster in a manner that increases service efficiency by detaching databases that are idle so as to reclaim cluster resources from the detached databases that are idle, but also reserving sufficient cluster resources to meet resources needed when detached databases are resumed, and wherein the instructions configure the computer system to perform at least the following:
identifying candidate databases that can be quiesced by detaching them so that cluster resources can be reclaimed, wherein the identification of the candidate databases is performed by identifying a plurality of quiescence policies, and wherein each of the quiescence policies define one or more conditions which when met, identify that cluster resources should be reclaimed from by detaching the candidate databases that meet the one or more conditions of one of the quiescence policies;
for each quiescence policy in the plurality of quiescence policies, identify an amount of reclaimed capacity defining the amount of cluster resources that will be reclaimed by detaching any database that meets the one or more conditions of the quiescence policy when implementing the quiescence policy;
for each quiescence policy in the plurality of quiescence policies, identify an amount of reserve capacity defining the amount of cluster resources that should be reserved in the event a detached database that meets the one or more conditions of the quiescence policy is resumed;
select a quiescence policy from among the plurality of quiescence policies that represents an attempt to optimize potential reclaimed cluster resources by using telemetry data collected from the cluster to model application of quiescence policies by comparing the reclaimed capacities and reserve capacities for each of the quiescence policies in the plurality of quiescence policies, and wherein the attempt to optimize comprises selecting a quiescence policy that attempts to maximize the reclaimed capacity of cluster resources for detached databases while minimizing the reserved capacity of cluster resources required for resumed databases; and
book cluster resources for cluster entities on the cluster according to the selected quiescence policy.

2. The computer system of claim 1 , wherein the telemetry data comprises recent real world telemetry data collected within a predetermined time frame from the cluster.

3. The computer system of claim 1 , wherein the telemetry data comprises recent real world telemetry data collected within a predetermined time frame from the cluster.

4. The computer system of claim 1 , wherein the telemetry data comprises the telemetry data comprises recent real world telemetry data collected within a predetermined time frame from the cluster and contextually relevant telemetry data collected from the cluster that is contextually relevant to a current or anticipated condition of the cluster.

5. The computer system of claim 1 , wherein the one or more computer-readable media further have stored thereon instructions that are executable by the one or more processors to configure the computer system to identify a resume cost for resuming each detached database, and wherein selecting a quiescence policy from among the plurality of quiescence policies is based on the resume cost for resuming each detached database.

6. The computer system of claim 1 , wherein the one or more computer-readable media further have stored thereon instructions that are executable by the one or more processors to configure the computer system to identify a capacity compensation cost that may be incurred for failing to meet a service level agreement for a resumed database, and wherein selecting a quiescence policy from among the plurality of quiescence policies is based on the compensation cost.

7. In a computing environment, a method of managing cluster resources for a cluster; in a manner that increases service efficiency by detaching databases that are idle so as to reclaim cluster resources from the detached databases that are idle, but also reserving sufficient cluster resources to meet resources needed when detached databases are resumed, and wherein the method comprises:
identifying candidate databases that can be quiesced by detaching them so that cluster resources can be reclaimed, wherein the identification of the candidate databases is performed by identifying a plurality of quiescence policies, and wherein each of the quiescence policies define one or more conditions which when met, identify that cluster resources should be reclaimed by detaching the candidate databases that meet the one or more conditions of one of the quiescence policies;
for each quiescence policy in the plurality of quiescence policies, identifying an amount of reclaimed capacity defining the amount of cluster resources that will be reclaimed by detaching any database that meets the one or more conditions of the quiescence policy when implementing the quiescence policy;
for each quiescence policy in the plurality of quiescence policies, identifying an amount of reserve capacity defining the amount of cluster resources that should be reserved in the event a detached database that meets the one or more conditions of the quiescence policy is resumed;
selecting a quiescence policy from among the plurality of quiescence policies that represents an attempt to optimize potential reclaimed cluster resources by using telemetry data collected from the cluster to model application of quiescence policies by comparing the reclaimed capacities and reserve capacities for each of the quiescence policies in the plurality of quiescence policies, and wherein the attempt to optimize comprises selecting a quiescence policy that attempts to maximize the reclaimed capacity of cluster resources for detached databases while minimizing the reserved capacity of cluster resources required for resumed databases; and
booking cluster resources for cluster entities on the cluster according to the selected quiescence policy.

8. The method of claim 7 , wherein the telemetry data comprises recent real world telemetry data collected within a predetermined time frame from the cluster.

9. The method of claim 7 , wherein the telemetry data comprises recent real world telemetry data collected within a predetermined time frame from the cluster.

10. The method of claim 7 , wherein the telemetry data comprises the telemetry data comprises recent real world telemetry data collected within a predetermined time frame from the cluster and contextually relevant telemetry data collected from the cluster that is contextually relevant to a current or anticipated condition of the cluster.

11. The method of claim 7 , further comprising identifying a resume cost for resuming each detached database, and wherein selecting a quiescence policy from among the plurality of quiescence policies is based on the resume cost for resuming each detached database.

12. The method of claim 7 , further comprising identifying a capacity compensation cost that may be incurred for failing to meet a service level agreement for a resumed database, and wherein selecting a quiescence policy from among the plurality of quiescence policies is based on the compensation cost.

13. A cluster system comprising:
a plurality of cluster resources;
a plurality of cluster entities deployed on the cluster resources; and
wherein the cluster system comprises one or more pr

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Ramachandra Karthik Saligrama
• Dewitt David Johns
• Xu Shize
• Guo Qun
• Kalhan Ajay
• Carlin Peter Alan
• Lang Willis

Classifications

• G06F11/3442

  for planning or managing the needed capacity · CPC title

• G06F9/5061

  Partitioning or combining of resources · CPC title

• G06F16/00

  Information retrieval; Database structures therefor; File system structures therefor · CPC title

• G06F9/5027

  the resource being a machine, e.g. CPUs, Servers, Terminals · CPC title

• H04L67/1025 Primary

  Dynamic adaptation of the criteria on which the server selection is based · CPC title