EXPRESSING FAULT CORRELATION CONSTRAINTS

• US 20100107015A1
• Filed: 10/24/2008
• Published: 04/29/2010
• Est. Priority Date: 10/24/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Bernabeu-Auban, Jose, Khalidi, Yousef A.
• Granted Patent: US 7,996,719 B2
• Time in Patent Office: Days
• US Class Current: 714/38
• CPC Class Codes: G06F 11/008 Reliability or availability analysis

Abstract

Methods, systems, and computer-readable media for expressing fault correlation constrains to a developer of a service application are provided. Initially, the fault correlation constraints are identified by inspecting an architecture of resources comprising a data center. Based on the resources, sets of nodes that become concurrently unavailable due to a failure mode or update scheme are aggregated. These fault correlation constraints, upon being communicated to the developer, then may be incorporated into a service model produced thereby. Accordingly, the service model is tailored to the specific characteristics of the data center and is provided with instructions as to how to partition instances of the service application throughout the distributed resources of the data center. Thus, the fault correlation constraints provide a developer with insight into rules and structures on which the data center is constructed and assurances of a certain level of fault tolerance upon deploying the service application.

Claims

1. One or more computer-readable media having computer-executable instructions embodied thereon that, when executed, performs a method for facilitating persistent availability of a service application by expressing one or more fault correlation constraints to a developer of the service application, the method comprising:

identifying fault domains within a hosting environment, wherein the fault domains each represent a set of nodes that each concurrently become unavailable upon an occurrence of a particular class of failures within the hosting environment;

expressing to the developer of the service application one or more fault correlation constraints wherein, the one or more correlation constraints include an abstraction of the identified fault domains;

receiving from the developer a service model that provides instructions for allocating one of the set of nodes within one or more of the fault domains within the hosting environment to execute at least one role of the service application, wherein the at least one role comprises a component program that supports the functionality of the service application; and

based on the instruction of the service model, initiating an instantiation process that places the one or more instances of the at least one role on the allocated one of the set of nodes.

2. The computer-readable media of claim 1, further comprising selecting the class of failures that each represent a failure mode of hardware or software components that, upon occurrence, induces a respective set of nodes to cease functioning properly.

3. The computer-readable media of claim 2, wherein selecting comprises:

constructing a hierarchy of classes of failures based on characteristics of the hosting environment, wherein the hierarchy is organized based on a frequency of occurrence of each of the classes of failures; and

utilizing the hierarchy for selecting one or more of the classes of failures for derivation of the fault domains.

4. The computer-readable media of claim 1, wherein the instructions of the service model comprise a specification for deployment that influences the instantiation process that places the one or more instances of the at least one role of the service application within the hosting environment.

5. The computer-readable media of claim 4, wherein the specification for deployment delimits a minimum number of the fault domains, per failure mode, to allocate to execute the one or more instances of the at least one role.

6. The computer-readable media of claim 5, wherein the specification for deployment defines which channels, established within the hosting environment, are utilized as communication paths between the placed one or more instances of differing roles once instantiated on the appropriate fault domains.

7. The computer-readable media of claim 1, wherein the instructions of the service model further comprise configuration settings to administer to the one or more instances during the instantiation process, wherein the configuration settings include information related to the allocated one or more of the fault domains on which the one or more instances are placed.

8. The computer-readable media of claim 1, wherein each of the set of nodes represents a computing device capable of running the one or more instances of the service application within the hosting environment.

9. A computerized method for developing a failure model of a data center for delivery to a developer of a service application, the method comprising:

identifying fault correlation constraints of a data center, wherein identifying comprises;

(a) inspecting an architecture of resources utilized by a data center to ascertain modes of failure of the data center;

(b) enumerating at least a first set of nodes within the data center that concurrently become unavailable upon an occurrence of the ascertained failure mode;

(c) enumerating at least a second set of nodes within the data center that are intentionally induced to become concurrently unavailable in accordance with an update scheme; and

(d) abstracting the fault correlation constraints from at least one the first set of nodes and the second set of nodes; and

at least temporarily storing the fault correlation constraints within a failure model that is deliverable to the developer of the service application, wherein the service application is executable upon being distributed throughout the architecture of resources of the data center.

10. The computerized method of claim 9, the method further comprising:

expressing, within the failure model, the enumerated first set of nodes as a fault domain and the enumerated second set of nodes as an update domain;

constructing semantics that influence the distribution of the service application throughout the data center and define a maximum aggregate number of fault domains and update domains that are allowed to be concurrently nonfunctional when executing the distributed service application; and

at least temporarily storing the semantics within the failure model.

11. The computerized method of claim 10, the method further comprising:

attaining a service model upon delivering the failure model to the developer of the service application, wherein instructions of the service model are derived from the fault correlation constraints and the semantics;

reading the service model to determine one or more roles that each represent a particular class of component of the service application; and

reading the service model to determine how many instances of each of the one or more roles to place within the data center,

wherein each of the instances is a replication of the particular class of component, and

wherein each of the instances of the particular class of components is partitioned at separate fault domains and update domains in accordance with the instructions of the service model.

12. The computerized method of claim 11, the method further comprising initiating an instantiation process for deploying the instances of the one or more roles of the service application in accordance with the service model, the instantiation process comprising:

recursively selecting the fault domains and the update domains of the data center that satisfy the service model;

installing the instances of the service application at the selected fault domains and the selected update domains; and

deriving from the service model configuration settings for administration to the installed instances.

13. The computerized method of claim 12, wherein the service model includes a deployment specification that manages the installation of the instances of each of the one or more roles, thereby distributing component programs of the service application, which support the functionality thereof, throughout the data center, wherein the deployment specification indicates a minimum number of the fault domains, associated with each of the failure modes, to assign to each of the one or more roles of the service application for placement of the instances of the one or more roles thereon.

14. The computerized method of claim 13, wherein the deployment specification indicates a minimum number of the update domains to assign to each of the one or more roles of the service application for placement of the instances of the one or more roles thereon.

15. The computerized method of claim 14, the method further comprising:

correlating each of the enumerated first set of nodes to one of the fault domains and correlating each of the enumerated second set of nodes to one of the update domains; and

utilizing the correlations to install the instances of each of the one or more roles at the appropriate first and second set of nodes in accordance with the deployment specification.

16. The computerized method of claim 15, the method further comprising mapping which of the fault domains and which of the update domains is assigned to each of the installed instances of each of the one or more roles.

17. The computerized method of claim 16, the method further comprising teaching the instances of the one or more roles to communicate therebetween across the assigned fault domains and the assigned update domains, wherein teaching comprises:

incorporating the mapping in the configuration settings;

automatically propagating the configuration settings to each of the installed instances; and

teaching each of the installed instances a location within the data center of other installed instances that support the service application by administering the configuration settings to the installed instances of the one or more roles.

18. The computerized method of claim 12, wherein recursively selecting the fault domains and the update domains of the data center that satisfy the service model comprises:

selecting one or more of the fault domains for installing the instances of a first role, of the one or more roles, of the service application, wherein the selected one or more fault domains comprises the first set of nodes;

selecting one or more of the update domains for installing the instances of the first role, wherein the selected one or more update domains comprises the second set of nodes; and

selecting a node of a third set of nodes for installation of one of the instances of the first role, wherein the node of the third set of nodes is supported by resources, utilized by a data center, that are uncommon to the resources underlying the first set of nodes and the second set of nodes.

19. A computer system for performing a method that abstracts fault domains from resources of a data center, the computer system comprising a processing unit coupled to a computer storage medium, the computer storage medium having stored thereon a plurality of computer software components executable by the processing unit, the computer software components comprising:

an abstraction component to abstract the fault domains within the data center by grouping together sets of nodes that concurrently become unavailable upon an occurrence of a mode of failure of the resources;

a communication component to communicate the abstracted fault domains and semantics, which influence distribution of a service application throughout the data center, as fault correlation constraints, wherein the semantics define a maximum aggregate number of fault domains and update domains that are allowed to be concurrently nonfunctional when executing the distributed service application, and to receive a service model that is based, in part, on the fault correlation constraints and that provides instructions for allocating nodes, within the sets of nodes encompassed by the fault domains, for accommodating instances of a role of the service application,

wherein the role comprises a component program that supports the functionality of the service application,

wherein each of the instances includes a single replication of the role, and

wherein the fault domains provide a platform for executing the instances of the role; and

a propagation component that satisfies the instructions of the service model by installing the instances on the allocated nodes and administering parameters to configure the installed instances, wherein the configuring includes directing each of the installed instances to a location of other installed instances within the data center.

20. The computer system of claim 19, wherein the abstraction component is further configured for identifying various classes of failures of the data center that cause sets of nodes to concurrently become unavailable, constructing a hierarchy of the classes of failures based on characteristics of the data center, wherein the hierarchy is organized based on a frequency of occurrence of each of the classes of failures, and utilizing the hierarchy to select one or more of the classes of failures with a high frequency of occurrence as the fault domains.