Diagnosis of application server performance problems via thread level pattern analysis

• US 9,442,817 B2
• Filed: 10/08/2010
• Issued: 09/13/2016
• Est. Priority Date: 10/08/2010
• Status: Active Grant

• Current Assignee: Vmware LLC (Broadcom, Inc.)
• Original Assignee: VMware, Inc. (Broadcom, Inc.)
• Inventors: Yu, Xuwen, Desai, Aalap, Dong, Aiguo, Liang, Xiaojun, Kambo, Rajit
• Primary Examiner(s): Chea, Philip
• Assistant Examiner(s): MCADAMS, BRAD
• Application Number:
  US12/901,228
• Publication Number:
  US 20120089724A1
• Time in Patent Office: 2,167 Days
• Field of Search: 709/224
• US Class Current: 1/1
• CPC Class Codes: G06F 11/328 Computer systems status display G06F11/327 takes precedence
• CPC Class Codes: G06F 11/3409 for performance assessment
• 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 2209/508 Monitor

Abstract

Performance of server that is managing a group of servers, such as in a cloud, is analyzed by identifying patterns of threads executed in the server. A command is issued, repeatedly over a time period, to an operating system of the server to return a resource consumption value representative of the amount of computing resources consumed by one or more threads. The returned resource consumption values are stored and used to generate points on a graph which depicts the amount of computing resources consumed by the one or more threads over the time period and provide task breakdowns and latency information for each of the tasks.

Claims

1. A method of monitoring performance of an application server for a group of servers in which one or more threads associated with multiple tasks that manage the group of servers are executed, said method comprising:

issuing a command to an operating system of the application server to return a resource consumption value representative of the amount of computing resource consumed by one or more threads, the resource consumption value including at least an identification of one of the threads and a corresponding amount of computing resource consumed by the one of the treads;

storing the resource consumption value;

repeating said issuing and said storing multiple times over a time period;

generating a graph of the amount of computing resources consumed by the one of the threads over the time period based on the stored resource consumption values;

identifying tasks in the one of the threads in the graph, including determining at least one task switch location on the graph based on a rate of change in resource consumption;

comparing the graph with predetermined patterns, wherein the predetermined patterns include a duplicate task pattern that indicates duplicate tasks defined by active periods of resource consumption separated by at least one inactive period of resource consumption; and

identifying potential performance improvement areas in the application server based on said comparing, including determining that the graph indicates at least two duplicate tasks based on whether the graph includes the duplicate task pattern; and

determining that the graph has the duplicate task pattern if two consecutive tasks represented in the graph as active periods of resource consumption separated by at least one inactive period of resource consumption have similar thread areas and similar latencies during their active periods of resource consumption, wherein the predetermined patterns include at least one of an inactive period pattern that includes an inactive period of an undesirable duration between two consecutive tasks represented in the pattern.

2. The method of claim 1, wherein the one of the threads includes multiple tasks and each task is associated with a distinct rising edge and a falling edge represented on the graph.

3. The method of claim 1, wherein said repeating is carried out at a predetermined rate that is user-configurable.

4. The method of claim 1, wherein the computing resource is one of CPU, system memory, and disk IO.

5. The method of claim 1, further comprising:

filtering threads so that they are not represented in the graph.

6. The method of claim 5, wherein the threads are filtered based on their thread area.

7. The method of claim 5, wherein the threads are filtered if they are background threads or system threads.

8. A method of identifying potential performance improvement areas in an application server for a group of servers in which one or more threads associated with multiple tasks that manage the group of servers are executed, said method comprising:

issuing a command to an operating system of the application server to return a resource consumption value representative of the amount of computing resource consumed by one or more threads, the resource consumption value including at least an identification of one of the threads and a corresponding amount of computing resource consumed by the one of the treads;

storing the resource consumption value;

repeating said issuing and said storing multiple times over a time period;

generating a graph of changes in the resource consumption value for the one of the threads over the time period based on the stored resource consumption values;

identifying tasks in the one of the threads in the graph, including determining at least one task switch location on the graph based on a rate of change in resource consumption;

comparing changes in the resource consumption value for the one of the threads over the time period with predetermined patterns, wherein the predetermined patterns include a duplicate task pattern that indicates duplicate tasks defined by active periods of resource consumption separated by at least one inactive period of resource consumption; and

identifying potential performance improvement areas in the application server based on said comparing, including determining that the graph indicates at least two duplicate tasks based on whether the graph includes the duplicate task pattern; and

determining that the graph has the duplicate task pattern if two consecutive tasks represented in the graph as active periods of resource consumption separated by at least one inactive period of resource consumption have similar thread areas and similar latencies during their active periods of resource consumption, wherein the predetermined patterns include at least one of an inactive period pattern that includes an inactive period of an undesirable duration between two consecutive tasks represented in the pattern.

9. A non-transitory computer readable storage medium comprising instructions to be executed by a processing unit of an application server for a group of servers in which one or more threads associated with multiple tasks that manage the group of servers are executed, wherein the instructions, when executed by the processing unit, cause the application server to carry out the steps of:

issuing a command to an operating system of the application server to return a resource consumption value representative of the amount of computing resource consumed by one or more threads, the resource consumption value including at least an identification of one of the threads and a corresponding amount of computing resource consumed by the one of the treads;

storing the resource consumption value;

repeating said issuing and said storing multiple times over a time period;

generating data points for a graph showing the amount of computing resource consumed by the one of the threads over the time period using the stored resource consumption values;

identifying tasks in the one of the threads in the graph, including determining at least one task switch location on the graph based on a rate of change in resource consumption;

comparing the graph with predetermined patterns, wherein the predetermined patterns include a duplicate task pattern that indicates duplicate tasks defined by active periods of resource consumption separated by at least one inactive period of resource consumption; and

identifying potential performance improvement areas in the application server based on said comparing, including determining that the graph indicates at least two duplicate tasks based on whether the graph includes the duplicate task pattern; and

determining that the graph has the duplicate task pattern if two consecutive tasks represented in the graph as active periods of resource consumption separated by at least one inactive period of resource consumption have similar thread areas and similar latencies during their active periods of resource consumption, wherein the predetermined patterns include at least one of an inactive period pattern that includes an inactive period of an undesirable duration between two consecutive tasks represented in the pattern.

10. The non-transitory computer readable storage medium of claim 9, wherein said repeating is carried out at a predetermined rate that is user-configurable.

11. The non-transitory computer readable storage medium of claim 9, wherein the computing resource is one of CPU, system memory, and disk IO.