Tuning and optimizing distributed systems with declarative models

• US 7,970,892 B2
• Filed: 06/29/2007
• Issued: 06/28/2011
• Est. Priority Date: 06/29/2007
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Panditharadhya, Prasad S., Borsa, Mariusz G., Zunino, Gilles C. J. A, Kulkarni, Amol S., Sedukhin, Igor, Eshner, Daniel
• Primary Examiner(s): Backer; Firmin
• Assistant Examiner(s): Chambers; Michael A
• Application Number:
  US11/771,827
• Publication Number:
  US 20090006063A1
• Time in Patent Office: 1,460 Days
• Field of Search: 709/224, 709/277, 709/203, 709/227, 709/220, 709/221, 707/100, 717/174, 715/733, 715/763
• US Class Current: 709/224
• CPC Class Codes: G06F 11/3447 Performance evaluation by modeling
• CPC Class Codes: G06F 11/3476 Data logging G06F11/14, G06F11/2205 take precedence
• CPC Class Codes: G06F 11/3495 for systems
• CPC Class Codes: G06F 2201/86 Event-based monitoring

Abstract

A system for automatically adjusting operation of a distributed application program includes analytics means having a monitoring component. The monitoring component receives one or more event streams of data corresponding to execution of one or more modules of the distributed application program. The monitoring component joins the received event streams with one or more declarative models to create operational data. A forensics component of the analytics means queries, such as by using data from a declarative model store, the operational data to identify trends or behavior information for the various modules or components of the distributed application program. A planner component then reviews data created by the forensics component to identify any needed changes to the declarative models. The planner component passes any modified declarative models back to the system, so that the distributed application program can operate more efficiently using the changes to the declarative models, as needed.

Claims

1. At a server configured in a computerized environment to manage one or more distributed application programs through one or more platform-specific drivers, a method of automatically monitoring and adjusting behaviors of a distributed application program through management of one or more declarative models, whereby the distributed application program automatically responds to fluctuations in hardware performance or user demand, comprising the acts of:

at a server system that includes implementation means;

sending an initial set of instructions corresponding to at least one declarative model of a distributed application program to at least one platform-specific driver of the implementation means, the at least one declarative model describing intent for operation of one or more modules of the distributed application program, and the at least one platform-specific driver providing instructions for executing the one or more modules as part of the distributed application program;

after initiation of execution of the distributed application program, receiving one or more event streams through the at least one platform-specific driver, the one or more event streams received through the at least one platform-specific driver containing information about real-time performance of the distributed application program during execution of the distributed application program within the computerized environment;

correlating the received execution information about performance of the distributed application program with the at least one declarative model, including the described operation of the one or more modules of the distributed application program; and

modifying the at least one declarative model in accordance with the received execution information, wherein the modification to the at least one declarative model adjusts one or more intended operations of one or more modules of the distributed application program.

2. The method as recited in claim 1, wherein the one or more event streams are received in response to a query that is scheduled to occur periodically.

3. The method as recited in claim 1, further comprising an act of a monitoring component creating operational data from the received one or more event streams, wherein the operational data includes behavior information metrics about execution of the distributed application program.

4. The method as recited in claim 3, wherein the act of creating operational data further comprises joining data of at least one of the one or more event streams with data in an declarative model repository.

5. The method as recited in claim 3, further comprising an act of a forensics component querying the operational data with one or more queries to identify one or more operational trends of the distributed application program.

6. The method as recited in claim 5, wherein at least one of the one or more queries includes one or more identifiers of a module in the distributed application program.

7. The method as recited in claim 5, wherein the one or more operational trends include an indication that one or more hardware components on which the distributed application program is executing are operating incorrectly or are failing.

8. The method as recited in claim 5, wherein the one or more operational trends include an indication that one or more software modules of the distributed application program are executing incorrectly or are failing.

9. The method as recited in claim 5, wherein the one or more operational trends include an indication of a frequency with which a machine that is executing at least a portion of the distributed application program is rebooting or otherwise restarting.

10. The method as recited in claim 5, further comprising an act of a planner component receiving the one or more operational trends from the forensics component.

11. The method as recited in claim 10, further comprising an act of the planner component correlating the one or more operational trends received from the forensics component with data of the at least one declarative model.

12. The method as recited in claim 11, wherein the act of modifying the at least one declarative model in accordance with the received execution information further comprises an act of the planner component modifying the at least one declarative model to create a modified declarative model.

13. The method as recited in claim 12, wherein the modified declarative model identifies a different set of one or more hardware components on which at least a portion of the distributed application program should be executed.

14. The method as recited in claim 12, wherein the modified declarative model reflects a change in execution or use policy for at least a portion of the distributed application program.

15. The method as recited in claim 12, further comprising an act of sending the modified declarative model to the implementation means, wherein information of the modified declarative model can be processed through the platform-specific driver.

16. The method as recited in claim 15, further comprising an act of the implementation means sending a subsequent set of instructions corresponding to the modified declarative model to the platform-specific driver.

17. At a server configured in a computerized environment to manage operations of one or more distributed application programs through one or more platform-specific drivers using one or more declarative models, a method of automatically adjusting operations of a distributed application program through periodic monitoring and tuning of the one or more declarative models during execution of the distributed application program, comprising:

at a server system comprising implementation means and analytics means;

implementing, via implementation means, an initial set of platform-specific instructions for a distributed application container on one or more different computer systems based on one or more declarative models for the distributed application program, the one or more declarative models describing behavior of one or more modules of the distributed application program, without describing how to implement such behavior of the one or more modules, the one or more declarative models loosely coupling components of an application system, and the one or more declarative models specifying at least a domain name where a module can be executed, and a plurality of permission sets that can be associated with execution of a module;

initiating execution of the distributed application program using the one or more modules, wherein initiating execution of the distributed application program includes using a connection protocol based on proximity of two or more modules, which connection protocol is not specified by the one or more declarative models;

after initiating execution of the distributed application program, monitoring real-time behavioral information for the one or more modules of the distributed application program at the one or more different computer systems, wherein monitoring real-time behavioral information is through one or more event streams received through at least one platform-specific driver;

determining from the monitored real-time behavioral information that one or more corresponding operational trends fall outside a desired parameter, including at least one operational trend indicating that a computer system on which at least one of the one or more modules is installed is rebooting too frequently;

automatically adjusting at least one declarative model to account for the computer system on which at least one of the one or more modules is installed rebooting too frequently; and

implementing, via the implementation means, a subsequent set of platform-specific instructions for the distributed application container based on the adjusted at least one declarative model.

18. The method as recited in claim 17, wherein the one or more operational trends that fall outside the desired parameter further indicate at least:

that at least a module of the distributed application program is failing too frequently.

19. The method as recited in claim 17, wherein the one or more operational trends that fall outside the desired parameter correspond to a usage spike and further indicate that at least one of the one or more modules is receiving a number of processing requests during a specified time interval that exceeds a parameter specified in the one or more declarative models.

20. At a server configured in a computerized environment to manage one or more distributed application programs through one or more platform-specific drivers, a computer program storage product having computer-executable instructions stored thereon that, when executed, cause one or more processors in the server to perform a method comprising the acts of:

sending an initial set of instructions corresponding to at least one declarative model of a distributed application program to at least one platform-specific driver of implementation means, the at least one declarative model describing intent for operation of one or more modules of the distributed application program, and the at least one platform-specific driver providing instructions for executing the one or more modules as part of the distributed application program;

after initiation of execution of the distributed application program, receiving one or more event streams through the at least one platform-specific driver, the one or more event streams received through the at least one platform-specific driver containing information about real-time performance of the distributed application program during execution of the distributed application program within the computerized environment;

correlating the received execution information about performance of the distributed application program with the at least one declarative model, including the described operation of the one or more modules of the distributed application program; and

modifying the at least one declarative model in accordance with the received execution information, wherein the modification to the at least one declarative model adjusts one or more intended operations of one or more modules of the distributed application program.