Sampling technique to adjust application sampling rate

Patent metadata

Field	Value
Publication number	US-10108522-B1
Application number	US-201715605574-A
Country	US
Kind code	B1
Filing date	May 25, 2017
Priority date	May 25, 2017
Publication date	Oct 23, 2018
Grant date	Oct 23, 2018

Abstract

One embodiment provides a method, including: identifying an application executing across a plurality of processors, wherein the application generates communication messages between the plurality of processors; collecting, using a profiler, a plurality of samples of the messages, wherein the plurality of samples are collected at a predetermined sampling rate and wherein the collecting comprises interrupting the execution of the application to collect a sample; determining, using an adaptive sampling technique, if the sampling rate comprises a sampling rate that is sufficiently low while meeting a predetermined quality threshold; adjusting, based upon the sampling rate not comprising a sampling rate that is sufficiently low, the sampling rate, and then collecting samples until the sampling rate comprises the sampling rate that is sufficiently low while meeting a predetermined quality threshold; and building, using the collected samples, an output performance profile of the application.

First claim

What is claimed is:

1. A method, comprising:
utilizing at least one processor to execute computer code that performs the steps of:
identifying an application executing across a plurality of processors, wherein the application generates communication messages between a subset of the plurality of processors during execution;
collecting, using a profiler, a plurality of samples of the communication messages, wherein the plurality of samples are collected at a sampling rate and wherein the collecting comprises interrupting the execution of the application to collect a sample;
determining, using an adaptive sampling technique, if the sampling rate comprises a sampling rate that is sufficiently low to reduce the interruptions to the execution of the application, while meeting a predetermined quality threshold, wherein the adaptive sampling techniques comprises a tensor adaptation technique comprising:
generating a tensor comprising the collected plurality of samples, the generated tensor comprises a plurality of missing values,
filling the plurality of missing values in the generated tensor by applying a tensor factorization technique to reconstruct the tensor, and
comparing a quality of the reconstructed tensor to the predetermined quality threshold;

adjusting, based upon the sampling rate not comprising a sampling rate that is sufficiently low to reduce the interruptions while meeting the predetermined quality threshold, the sampling rate, and then collecting samples until the sampling rate comprises the sampling rate that is sufficiently low while meeting the predetermined quality threshold; and
building, using the collected plurality of samples and the samples collected using the adjusted sampling rate, an output performance profile of the application.

2. The method of claim 1 , wherein the adjusting the sampling rate and collecting samples comprises iteratively performing the steps of: the collecting a plurality of samples, the determining, and the adjusting the sampling rate.

3. The method of claim 1 , wherein the adaptive sampling technique comprises a lightweight adaptive sampling technique comprising:
using a message predictor to make a plurality of predictions regarding information of a plurality of future messages; and
determining an prediction rate associated with the plurality of predictions by verifying, after the future message has been sent, the predicted information regarding the future message.

4. The method of claim 3 , wherein the lightweight adaptive sampling technique comprises classifying the application as regular based upon determining the prediction rate is above a predetermined threshold; and
wherein the adjusting the sampling rate comprises lowering the sampling rate based upon the application being classified as regular.

5. The method of claim 3 , wherein the lightweight adaptive sampling technique comprises classifying the application as irregular based upon determining the prediction rate is below a predetermined threshold; and
wherein the adjusting the sampling rate comprises raising the sampling rate based upon the application being classified as irregular.

6. The method of claim 1 , wherein the tensor adaptation technique comprises determining the quality of reconstruction is below the predetermined quality threshold; and
wherein the adjusting the sampling rate comprises increasing the sampling rate.

7. The method of claim 1 , wherein the tensor adaptation technique comprises determining the quality of reconstruction is above the predetermined quality threshold; and
wherein the adjusting the sampling rate comprises decreasing the sampling rate.

8. The method of claim 1 , wherein the building an output performance profile of the application comprises stitching together (i) the collected plurality of samples and (ii) the samples collected using the adjusted sampling rate.

9. The method of claim 1 , wherein the determining and the adjusting the sampling rate are triggered when the application transfers to a new function within the application.

10. An apparatus, comprising:
at least one processor; and
a computer readable storage medium having computer readable program code embodied therewith and executable by the at least one processor, the computer readable program code comprising:
computer readable program code that identifies an application executing across a plurality of processors, wherein the application generates communication messages between a subset of the plurality of processors during execution;
computer readable program code that collects, using a profiler, a plurality of samples of the communication messages, wherein the plurality of samples are collected at a sampling rate and wherein the collecting comprises interrupting the execution of the application to collect a sample;
computer readable program code that determines, using an adaptive sampling technique, if the sampling rate comprises a sampling rate that is sufficiently low to reduce the interruptions to the execution of the application, while meeting a predetermined quality threshold, wherein the adaptive sampling techniques comprising a tensor adaptation technique comprising:
generating a tensor comprising the collected plurality of samples, the generated tensor comprises a plurality of missing values,
filling the plurality of missing values in the generated tensor by applying a tensor factorization technique to reconstruct the tensor, and
comparing a quality of the reconstructed tensor to the predetermined quality threshold;

computer readable program code that adjusts, based upon the sampling rate not comprising a sampling rate that is sufficiently low to reduce the interruptions while meeting the predetermined quality threshold, the sampling rate, and then collecting samples until the sampling rate comprises the sampling rate that is sufficiently low while meeting the predetermined quality threshold; and
computer readable program code that builds, using the collected plurality of samples and the samples collected using the adjusted sampling rate, an output performance profile of the application.

11. A computer program product, comprising:
a computer readable storage medium having computer readable program code embodied therewith, the computer readable program code executable by a processor and comprising:
computer readable program code that identifies an application executing across a plurality of processors, wherein the application generates communication messages between a subset of the plurality of processors during execution;
computer readable program code that collects, using a profiler, a plurality of samples of the communication messages, wherein the plurality of samples are collected at a sampling rate and wherein the collecting comprises interrupting the execution of the application to collect a sample;
computer readable program code that determines, using an adaptive sampling technique, if the sampling rate comprises a sampling rate that is sufficiently low to reduce the interruptions to the execution of the application, while meeting a predetermined quality threshold, wherein the adaptive sampling techniques comprises a tensor adaptation technique comprising:
generating a tensor comprising the collected plurality of samples, the generated tensor comprises a plurality of missing values,
filling the plurality of missing values in the generated tensor by applying a tensor factorization technique to reconstruct the tensor, and
comparing a quality of the reconstructed tensor to the predetermined quality threshold;

computer readable program code that adjusts, based upon the sampling rate not comprising a sampling rate that is sufficiently low to reduce t

Assignees

• IBM

Inventors

• Jain Nitisha
• Murali Prakash

Classifications

• G06F11/302

  where the computing system component is a software system · CPC title

• G06F11/3409 Primary

  for performance assessment · CPC title

• G06F11/3604

  Analysis of software for verifying properties of programs (testing of software G06F11/3668) · CPC title

• G06F2201/865

  Monitoring of software · CPC title