Resource Consumption Estimation with Out of Norm Determination in Mobile Devices

• US 20150011177A1
• Filed: 07/02/2013
• Published: 01/08/2015
• Est. Priority Date: 07/02/2013
• Status: Active Grant

• Current Assignee: AT&T Mobility IP LLC (AT&T, Inc.)
• Original Assignee: Carrier Iq
• Inventors: Gutierrez, Alberto Jr., Wang, Chang Jia
• Granted Patent: US 9,986,450 B2
• Time in Patent Office: Days
• US Class Current: 455/405
• CPC Class Codes: H04W 24/08 Testing, supervising or monitoring using real traffic
• CPC Class Codes: H04W 4/60 Subscription-based services using application servers or record carriers, e.g. SIM application toolkits

Abstract

A method estimates resource consumption between applications within and among a population of wireless mobile devices. Aggregating and analyzing samples for each member of the population of devices provides comparable individual results. Subsequently aggregating and analyzing only summary information for each member of the population of devices provides cluster, crowd, and global comparisons. An apparatus determines out of norm behavior of an individual within a crowd, or relative differences between two crowds.

Claims

1. A method to estimate if a first application substantially differs in resource consumption from a second application, the method comprising:

receiving, from each of a plurality of devices, a report of remaining resource availability measured at certain events in time;

receiving, recorded at each event-time, a list of applications that have been active since the last measurement;

for each first application on each device, estimating an average consumption for each first application;

determining a mean, median, standard deviation of resource consumption for each resource of each of a subset of devices which have a selected first application in the list of applications that have been active since the last measurement;

selecting the group of all devices in the population which have a selected first application in the list of applications;

determining a mean, median, standard deviation of resource consumption for all devices in the population which have a selected first application in its respective list of applications recorded at each event-time; and

ascribing the mean, median, standard deviation of resource consumption to the selected first application.

2. The method of claim 1, further comprising:

for each second application on each device, estimating an average consumption for each second application;

determining a mean, median, standard deviation of resource consumption for each resource of each of a subset of devices which have a selected second application in the list of applications that have been active since the last measurement;

selecting the group of all devices in the population which has a selected second application in the list of applications;

determining a mean, median, standard deviation of resource consumption for all devices in the population which have a selected second application in its respective list of applications recorded at each event-time; and

ascribing the mean, median, standard deviation of resource consumption to the selected second application.

3. The method of claim 2, further comprising:

for each first or second application version on each device, estimating an average consumption for each first or second application version;

determining a mean, median, standard deviation of resource consumption for each resource of each of a subset of devices which have either selected first application or selected second application in the list of applications that have been active since the last measurement;

selecting the group of all devices in the population which has either one of or both selected first application and selected second application version in the list of applications;

determining a mean, median, standard deviation of resource consumption for all devices in the population which have one of or both first application and second application in its respective list of applications recorded at each event-time;

ascribing the mean, median, standard deviation of resource consumption to the set of first and second applications; and

determining if the differences among mean, median, standard deviations of resource consumption of first selected application and second selected application are statistically significant.

4. The method of claim 1 wherein the first application is a first version of a first application and the second application is a second version of the first application, whereby regression testing may determine that a newer version is an improvement or not an unintentional substantial degradation of a previous version.

5. A method to identify Apps that over consume resources on all devices at a specified level of confidence comprising:

receiving the results of a resource estimation;

testing if App X is in the top P-th percentile (e.g., 10%) of battery drain rate on all devices with 95% confidence;

generating a list of apps (for the specified crowd) that over consume resources. The exact form of statistical tests is a design choice. The figure illustrates a

6. The method of claim 6 wherein the test is a 2-sample t-test since the key test is comparing two distributions.

7. A method to find App-Device combinations where an App over consumes resources comprising:

testing an App against the statistics of all devices that exercise the App (in the given crowd);

testing an APP against a global set of apps and devices;

denoting an App-Device combination as high when its consumption (e.g., battery drain rate) is in the top P-th percentile (10%) as compared to the statistics of the global set (the crowd);

generating a list of App-Device combinations where the App over consumes resources.

8. The method of claim 7 wherein the test is 1-sample t-test because it is comparing a device mean to a distribution.