Analysis and visualization of cluster resource utilization

• US 8,990,551 B2
• Filed: 09/16/2010
• Issued: 03/24/2015
• Est. Priority Date: 09/16/2010
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Shafi, Hazim
• Primary Examiner(s): Abbaszadeh, Jaweed A
• Assistant Examiner(s): Chen, Xuxing
• Application Number:
  US12/883,859
• Publication Number:
  US 20120072758A1
• Time in Patent Office: 1,650 Days
• Field of Search: 713/100, 718/100
• US Class Current: 713/100
• CPC Class Codes: G06F 11/323 Visualisation of programs or trace data
• CPC Class Codes: G06F 11/3404 for parallel or distributed programming
• CPC Class Codes: G06F 11/3612 by runtime analysis performance monitoring G06F11/3466
• CPC Class Codes: G06F 2209/508 Monitor
• CPC Class Codes: G06F 9/5011 the resources being hardware resources other than CPUs, Servers and Terminals

Abstract

An analysis and visualization depicts how an application is leveraging processor cores of a distributed computing system, such as a computer cluster, in time. The analysis and visualization enables a developer to readily identify the degree of concurrency exploited by an application at runtime and the amount of overhead used by libraries or middleware. Information regarding processes or threads running on the nodes over time is received, analyzed, and presented to indicate portions of computer cluster that are used by the application, idle, other processes, and libraries in the system. The analysis and visualization can help a developer understand or confirm contention for or under-utilization of system resources for the application and libraries.

Claims

1. A method, comprising:

receiving a representative number of at least an amount of available cores in nodes of a distributed computing system;

receiving a plurality of event provider records including context switch records and runtime provider records, wherein context switch records are analyzed to determine a representative number of the cores used to perform at least one selected process over a course of time, and wherein the runtime provider records are analyzed to determine a representative number of the cores used to perform at least one library function that provides for processing the at least one selected process in the nodes over the course of time;

determining from the context switch records and the runtime provider records a duration of time each core of available cores performs the at least one selected process and the at least one selected library function over the course of time;

separating the course of time into periodic intervals;

determining a representative number of cores executing the selected process and library function for each of the periodic intervals on each node in the system; and

presenting the representative number of cores executing the selected process and library function for each of the periodic intervals against the representative number of at least the amount of available nodes over the course of time.

2. The method of claim 1 wherein receiving a representative number of at least the amount of available cores includes querying a system scheduler.

3. The method of claim 2 where the amount of available cores includes less than a number of total cores in the distributed computing system.

4. The method of claim 1 wherein the distributed computer system is a Beowulf Cluster.

5. The method of claim 1 wherein computing devices in the distributed computer system include more than one node.

6. The method of claim 1 wherein library functions are Message Passing Interface (MPI) functions.

7. The method of claim 1 wherein the context switch records include context switch events and the runtime provider records include runtime provider events.

8. The method of claim 7 wherein the context switch events are representative of threads executing in the selected process and the runtime provider events are representative of threads executing the at least one library function.

9. The method of claim 1 wherein the representative number of cores includes the average number of concurrently executing threads in the selected process over the course of the periodic interval and are combined to provide a representative number of nodes executing the selected process over the course of the periodic interval.

10. The method of claim 1 where in the at least one selected processes includes a process of interest, an idle process, and a system process.

11. The method of claim 1 and further comprising modifying clock information for at least some of the nodes in the distributed computing system to synchronize the occurrence of each event provider record across the nodes in the cluster.

12. The method of claim 1 wherein the representative number of cores for each of the periodic intervals against the representative number of at least the amount of available cores in the distributed computing system over the course of time is presented in an area graph.

13. The method of claim 1 and further including instrumenting runtime providers into library functions of interest.

14. A computer readable storage device storing computer executable instructions for controlling a computing device to perform a method comprising:

receiving a representative number of at least the amount of available cores in nodes of a distributed computing system, wherein each node includes at least one core;

receiving a plurality of event provider records including context switch records and runtime provider records, wherein context switch records include a representative number of the cores used to perform at least one selected process over a course of time, and wherein the runtime provider records include a representative number of the cores used to perform at least one library function that provides for processing the at least one selected process in the nodes over the course of time;

determining from the context switch records and the runtime provider records a duration of time each core of available cores performs the at least one selected process and the at least one selected library function over the course of time;

separating the course of time into periodic intervals;

determining a representative number of cores executing the selected process and library function for each of the periodic intervals; and

presenting the representative number of cores executing the selected process and library function for each of the periodic intervals against the representative number of at least the amount of available cores over the course of time.

15. The computer readable storage device of claim 14 wherein the amount of available cores in a distributed computing system are received by interrogating an operating system on the distributed computing system.

16. The computer readable storage device of claim 14 wherein the at least one library function includes one of Message Passing Interface (MPI) functions or Parallel Virtual Machine (PVM) functions.

17. The computer readable device of claim 14 wherein the at least one library function includes Send( ) and Recv( ) functions.

18. The computer readable storage device of claim 14 further including an integrated development environment.

19. A computer readable storage device storing a development tool including computer executable instructions for controlling a computing device, the development tool configured to perform a method comprising:

receiving a representative number of at least the amount of available nodes in a computer cluster computing system;

receiving a plurality of event provider records including context switch records and runtime provider records, wherein context switch records include a representative number of the nodes used to perform at least one selected process over a course of time, and wherein the runtime provider records include a representative number of the nodes used to perform input or output library functions that provides for processing the at least one selected process in the nodes over the course of time;

determining from the context switch records and the runtime provider records a duration of time each core of available cores performs the at least one selected process and the at least one selected library function over the course of time;

modifying clock information for at least some of the nodes in the distributed computing system to synchronize the occurrence of each event provider record;

separating the course of time into generally equal periodic intervals;

determining a representative number of nodes executing the selected process and a representative number of library function for each of the periodic intervals; and

presenting the each of the representative number of nodes executing the selected process and the representative number of library function for each of the periodic intervals against the representative number of at least the amount of available nodes over the course of time in an area graph.

20. The computer readable storage device of claim 19 wherein each node is a logical core of the computer cluster.