System and method for managing a computing cluster

• US 9,459,925 B2
• Filed: 06/23/2011
• Issued: 10/04/2016
• Est. Priority Date: 06/23/2010
• Status: Active Grant

• Current Assignee: Twilio, Inc.
• Original Assignee: Twilio, Inc.
• Inventors: Lawson, Jeffrey G., Wolthuis, John, Cooke, Evan Mansfield
• Primary Examiner(s): Chang, Tom Y
• Application Number:
  US13/167,562
• Publication Number:
  US 20120209968A1
• Time in Patent Office: 1,930 Days
• Field of Search: 709/220
• US Class Current: 1/1
• CPC Class Codes: G06F 9/5072 Grid computing
• CPC Class Codes: H04L 67/142 Managing session states for stateless protocols; Signalling session states; State transitions; Keeping-state mechanisms
• CPC Class Codes: H04M 3/22 Arrangements for supervision, monitoring or testing
• CPC Class Codes: H04M 7/00 Arrangements for interconnection between switching centres

Abstract

A method and system for managing a computing cluster including hosting a plurality of machines in a networked computing cluster, wherein the plurality of machines include service instances running on hosts, where the services have configured machine state; and wherein machine state includes configuration data and software of the machine; integrating the plurality of machines of the cluster with at least one configuration controller component; and the at least one configuration controller component, altering the machine state of at least one service instance.

Claims

1. A method for managing a telephony platform networked computing cluster hosted at a third party Internet-based multi-tenant computing platform, the method comprising:

at a configuration controller that is communicatively coupled to the third party Internet-based multi-tenant computing platform via the Internet;

the configuration controller establishing a configuration controller publication channel with a message broker that is constructed to provide machine state via the Internet to a machine that is subscribed to the configuration controller publication channel, the machine being a machine of a plurality of machines of the computing platform, the message broker being constructed to provide machine state notifications received from the machine to the configuration controller, the message broker receiving the machine state notifications via a machine publication channel between the machine and the message broker, establishing the configuration controller publication channel comprising;

the configuration controller storing a security key of the machine that is supplied at boot time of the machine and provided to the configuration controller via the message broker, the security key being used to securely identify the machine;

responsive to a first instruction received by the configuration controller from an outside system via an application programming interface (API) of the configuration controller, the configuration controller;

encrypting first telephony platform computing cluster software with the stored security key, and

providing the encrypted first telephony platform computing cluster software to the machine via the configuration controller publication channel, the first telephony platform computing cluster software including software for a first telephony platform service that includes a plurality of telephony call router services, a plurality of dedicated telephony service load balancer services, and a plurality of dependent telephony services, each call router service being communicatively coupled with a dedicated telephony service load balancer service, each load balancer service distributing telephony jobs of the associated telephony call router service to a plurality of dependent telephony services that are managed by the load balancer service, the telephony services including two or more of a media processor service, a routing service, a call queue service, a resource database service, and a proxy service;

the configuration controller establishing at least one configuration controller subscription with the message broker to receive machine state notifications of the machine via the Internet, wherein responsive to the machine decrypting the encrypted first telephony platform computing cluster software by using the security key, the machine executes the first telephony platform computing cluster software, and wherein the first telephony platform computing cluster software is constructed to control the machine to;

provide at least a portion of the first telephony platform service,

subscribe to new machine state published by the configuration controller via the configuration controller publication channel, and

publish a machine state notification of the machine to the subscribed configuration controller via the Internet responsive to a change in a machine state of the machine;

at the configuration controller, responsive to a published machine state notification of the machine, the notification being received via message broker;

updating a machine state database with the machine state notification,

calculating new machine state from the machine state notification and machine state stored in the machine state database, the machine state stored in the machine state database including machine state of the plurality of telephony call router services, telephony service load balancer services, and dependent telephony services of the computing cluster,

selectively publishing the calculated new machine state to the machine by using the security key, and

in a case where the notification is published by a dependent telephony service;

identifying the associated telephony service load balancer service by using the machine state database,

calculating new machine state for the identified load balancer service, and

publishing the calculated new machine state for the load balancer service to a machine of the load balancer service via the configuration controller publication channel, and

responsive to a second instruction received by the configuration controller from the outside system via the API of the configuration controller, the configuration controller;

encrypting second telephony platform computing cluster software with the stored security key, and

providing the encrypted second telephony platform computing cluster software to the machine via the configuration controller publication channel to replace the first telephony platform computing cluster software at the machine, the second telephony platform computing cluster software including software for a second telephony platform service that is different from the first telephony platform service, the machine decrypting the encrypted second telephony platform computing cluster software by using the security key and the machine replacing execution of the first telephony platform computing cluster software with execution of the second telephony platform computing cluster software to provide at least a portion of the second telephony platform service, the first telephony platform service being one of a plurality of telephony platform services including a voice communication service, a messaging communication service, and a video communication service, the second telephony platform service being one of the plurality of telephony platform services that is different from the first telephony platform service,

wherein in a case where the configuration controller selectively publishes the calculated new machine state to a machine of a dependent telephony service, the configuration controller publishes the calculated new machine state via a machine of the associated telephony service load balancer service, and

wherein machine state of a telephony service load balancer service identifies dependent telephony services of the load balancer,

wherein new machine state published by the configuration controller includes at least one of a telephony platform configuration data change, and a telephony platform software change,

wherein the configuration controller provides the first telephony platform computing cluster software to the machine via the Internet, receives machine state notifications from the computing platform via the Internet, and publishes new machine state to the computing platform via the Internet, and

wherein the outside system is an external system of a user of the computing platform.

2. The method of claim 1, further comprising sharing common services of at least two input channel types.

3. The method of claim 1, wherein calculating new machine state comprises the configuration controller running a finite state machine to determine a new machine state.

4. The method of claim 1, wherein updating a machine state database with the machine state notification comprises storing version history of machine state in a content delivery network, and further comprising the machines of the cluster pulling new versions of machine state from the content delivery network.

5. The method of claim 1, further comprising in a case where a service subscribed to new machine states of the configuration controller is processing a telephony session, publishing the calculated new machine state to the service after the telephony session ends.

6. The method of claim 5, further comprising in a case where a service subscribed to new machine states of the configuration controller is processing a telephony session, publishing the calculated new machine state to dependent services of the service.

7. The method of claim 1, wherein machine state includes dependent services information;

and wherein calculating new machine state from the machine state notification and machine state stored in the machine state database further comprises calculating a new machine state of a dependent service following a dependency graph formed from dependent services information in the machine state database.

8. The method of claim 1, further comprising, at a call router service, receiving calls and executing a telephony application during a telephony session.

9. The method of claim 1, further comprising, at a call router service, receiving telephony messages and executing telephony applications.

10. The method of claim 1, wherein each telephony call router service, load balancer service, and dependent telephony service has a configured machine state, and wherein the configured machine state includes service status, service configuration data, software of the service, and dependent service information.

11. The method of claim 10, further comprising, at the configuration controller, responsive to a published machine state notification of a service to which the configuration controller is subscribed, identifying service instances that are dependent on the service of the published machine state notification and building a dependency graph of services, wherein selectively publishing the calculated new machine state comprises publishing to subscribed services in an order based on the dependency graph.

12. The method of claim 10,

wherein the configuration controller includes an application programming interface (API),

the method further comprising:

receiving a service update message via the API of the configuration controller, the service update message specifying new machine state for at least one specified service of the computing cluster, the at least one specified service including at least one of a telephony call router service, a dedicated telephony service load balancer service, and a dependent telephony service of the computing cluster, and

providing the new machine state of the service update message to the at least one specified service.

13. The method of claim 12, wherein the API of the configuration controller is a REST API, and wherein the service update message is provided by a system that is external to the telephony platform computing cluster system.

14. The method of claim 13, further comprising verifying authenticity of published machine state with a shared key.

15. The method of claim 14, further comprising the services of the computing cluster maintaining at least an instance subscription, a service type subscription, and a status subscription to the configuration controller;

and selectively publishing the calculated new machine state comprises;

publishing to a service instance through an instance subscription;

publishing to a service of a first type through a service type subscription; and

publishing to service of a first status through a status subscription.

16. The method of claim 13, wherein new machine state published by the configuration controller includes at least one of a configuration data change, and a software change.

17. The method of claim 13, wherein new machine state provided by the configuration controller includes state of a new telephony service, and wherein the configuration controller changes a telephony service of at least one machine of the computing cluster to a new telephony service by publishing new machine state to the at least one machine, wherein each telephony service includes at least one of a telephony call router service, a dedicated telephony service load balancer service, and a dependent telephony service.

18. The method of claim 13,

wherein the configuration controller changes a telephony service of at least one machine of the computing cluster by publishing new machine state to the at least one machine,

wherein each telephony service includes at least one of a telephony call router service, a dedicated telephony service load balancer service, and a dependent telephony service,

wherein the configuration controller changes the telephony service of the at least one machine according to a type of input channel associated with the telephony service,

wherein the type of input channel includes at least one of voice, video, and messaging.

19. The method of claim 1, wherein the first telephony platform computing cluster software is replaced during operation of the first telephony platform service.