Techniques for deploying changes to improve reliability of a cloud service

Patent metadata

Field	Value
Publication number	US-12248816-B2
Application number	US-202217573481-A
Country	US
Kind code	B2
Filing date	Jan 11, 2022
Priority date	Jan 11, 2022
Publication date	Mar 11, 2025
Grant date	Mar 11, 2025

Abstract

A data processing system implements obtaining a set of input parameters associated with an update to be deployed to a plurality of server farms of a cloud-based service, wherein each server farm includes a primary replica configured to handle user traffic and a disaster recovery replica configured to handle user traffic responsive to a failure of the primary replica; determining temperature information for each of the server farms, ranking the server farms based on the temperature information to determine an order in which an update is to be deployed to the server farms; iteratively deploying the updates to the primary replicas of the server farms according to the ranking until an deployment threshold has been satisfied; and iteratively deploying the updates to the primary replicas of server farms for which the primary replicas have not yet been updated and to the disaster recovery replicas of the server farms.

First claim

What is claimed is:

1. A data processing system comprising:
a processor; and
a machine-readable medium storing executable instructions that, when executed, cause the processor to perform operations comprising:
receiving a set of input parameters associated with an update to be deployed to a plurality of server farms of a cloud-based service during a deployment, each server farm including a primary replica configured to handle user traffic and a disaster recovery replica configured to handle the user traffic upon a failure of the primary replica, the set of input parameters including a deployment threshold indicating an optimal number of the plurality of server farms for which the primary replica is updated prior to updating the disaster recovery replica of the plurality of server farms, and a deployment increment indicating a number of server farms to be updated during each phase of a plurality of phases of the deployment, wherein the optimal number is greater than zero and less than a total number of primary replicas in the plurality of server farms;
storing temperature information for each of the plurality of server farms, the temperature information including a first temperature associated with the primary replica and a second temperature associated with the disaster recovery replica, the first temperature representing a load on the primary replica and the second temperature representing a load on the disaster recovery replica;
iteratively deploying the update to the primary replica of the plurality of server farms during each phase of the deployment according to a ranking of the plurality of server farms and the deployment increment until the deployment threshold has been satisfied, wherein the ranking is based on the temperature information; and
after the deployment threshold has been satisfied, iteratively deploying the update to remaining primary replicas of server farms for which the primary replica has not yet been updated along with the disaster recovery replica of the plurality of server farms during subsequent phases of the deployment according to the ranking of the server farms and the deployment increment.

2. The data processing system of claim 1 , wherein the temperature information is updated periodically and the deployment increment is changed after the deployment threshold has been satisfied.

3. The data processing system of claim 2 , wherein the set of input parameters further include a deployment interval indicating a period to wait between each phase of the deployment.

4. The data processing system of claim 3 , the machine-readable medium storing executable instructions that, when executed, further cause the processor to perform operations comprising:
providing build information as an input to a machine learning model configured to output the deployment threshold, the deployment increment, and the deployment interval, the machine learning model being trained to predict the deployment threshold, the deployment increment, and the deployment interval based on training data representing previous updates deployed to the cloud-based service; and
receiving the set of input parameters including the deployment threshold, the deployment increment, and the deployment interval from the machine learning model.

5. The data processing system of claim 3 , wherein the plurality of server farms having primary replicas with a lower temperature are higher in the ranking than those of server farms having primary replicas with a higher temperature.

6. The data processing system of claim 3 , the machine-readable medium storing executable instructions that, when executed, further cause the processor to perform operations comprising:
after all primary replicas and all disaster recovery replicas have been updated, generating a report and sending the report to one or more administrators indicating that all of the primary replicas and all of the disaster recovery replicas have been updated.

7. A method implemented in a data processing system for deploying updates to a cloud-based service, the method comprising:
receiving a set of input parameters associated with an update to be deployed to a plurality of server farms of a cloud-based service during a deployment, each server farm including a primary replica configured to handle user traffic and a disaster recovery replica configured to handle the user traffic upon a failure of the primary replica, the set of input parameters including a deployment threshold indicating an optimal number of the plurality of server farms for which the primary replica is updated prior to updating the disaster recovery replica of the plurality of server farms, and a deployment increment indicating a number of server farms to be updated during each phase of a plurality of phases of the deployment, wherein the optimal number is greater than zero and less than a total number of primary replicas in the plurality of server farms;
storing temperature information for each of the plurality of server farms, the temperature information including a first temperature associated with the primary replica and a second temperature associated with the disaster recovery replica, the first temperature representing a load on the primary replica and the second temperature representing a load on the disaster recovery replica;
iteratively deploying the update to the primary replica of the plurality of server farms during each phase of the deployment according to a ranking of the plurality of server farms and the deployment increment until the deployment threshold has been satisfied, wherein the ranking is based on the temperature information; and
after the deployment threshold has been satisfied, iteratively deploying the update to remaining primary replicas of server farms for which the primary replica has not yet been updated along with the disaster recovery replica of the plurality of server farms during subsequent phases of the deployment according to the ranking of the server farms and the deployment increment.

8. The method of claim 7 , wherein the temperature information is updated periodically and the deployment increment is changed after the deployment threshold has been satisfied.

9. The method of claim 8 , wherein the set of input parameters further include a deployment interval indicating a period to wait between each phase of the deployment.

10. The method of claim 9 , further comprising:
providing build information as an input to a machine learning model configured to output the deployment threshold, the deployment increment, and the deployment interval, the machine learning model being trained to predict the deployment threshold, the deployment increment, and the deployment interval based on training data representing previous updates deployed to the cloud-based service; and
receiving the set of input parameters including the deployment threshold, the deployment increment, and the deployment interval from the machine learning model.

11. The method of claim 7 , wherein
the plurality of server farms having primary replicas with a lower temperature are higher in the ranking than those of server farms having primary replicas with a higher temperature.

12. The method of claim 8 , further comprising:
after all primary replicas and all disaster recovery replicas have been updated, generating a report and sending the report to one or more administrators indicating that all primary replicas and all disaster recovery replicas have been updated.

13. A machine-readable medium on which are stored instructions that, when executed, cause a processor of a programmable device to perform operations of:
receiving a set of input parameters associated wit

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Verma Nidhi
• Nigam Rahul
• Khanna Rohan

Classifications

• G06F9/5038

  considering the execution order of a plurality of tasks, e.g. taking priority or time dependency constraints into consideration (scheduling strategies G06F9/4881 and subgroups) · CPC title

• G06F8/656

  while running · CPC title

• H04L41/0654

  using network fault recovery (ring fault isolation or reconfiguration in loop networks without recovery actions by a network management system H04L12/437) · CPC title

• H04L67/1097

  for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS] · CPC title

• H04L67/1036

  Load balancing of requests to servers for services different from user content provisioning, e.g. load balancing across domain name servers · CPC title