4.2 Experimental Methodology
4.2.4 Experimental plan
We defined an experimental plan by considering different combinations of the levels and factors presented in the previous subsection. In turn, the experimental plan can be divided in three sets. The full experimental plan includes 72 experiments, for a total of 18 days of experimentation. All the experiments are listed in Table4.2and sorted from the oldest to the newest Android version. The three test plans are blocked full-factorial designs, in which one factor is fixed (i.e., the Android device or version, according our research questions), while we vary all the other parameters.
TABLE4.2: Experimental plan of the case study
ID DEV VER APP EVENTS STO
EXP1 HUAWEIP8 ANDROID5 EU MIXED1 NORMAL
EXP2 HUAWEIP8 ANDROID5 EU MIXED1 FULL
EXP3 HUAWEIP8 ANDROID5 EU MIXED2 NORMAL
EXP4 HUAWEIP8 ANDROID5 EU MIXED2 FULL
EXP5 HUAWEIP8 ANDROID5 EU MIXED3 NORMAL
EXP6 HUAWEIP8 ANDROID5 EU MIXED3 FULL
EXP7 HUAWEIP8 ANDROID5 CHINA MIXED1 NORMAL
EXP8 HUAWEIP8 ANDROID5 CHINA MIXED1 FULL
EXP9 HUAWEIP8 ANDROID5 CHINA MIXED2 NORMAL
EXP10 HUAWEIP8 ANDROID5 CHINA MIXED2 FULL
Table 4.2: Experimental plan of the case study – continued from previous page
ID DEV VER APP EVENTS STO
EXP11 HUAWEIP8 ANDROID5 CHINA MIXED3 NORMAL
EXP12 HUAWEIP8 ANDROID5 CHINA MIXED3 FULL
EXP13 HUAWEIP8 ANDROID6 EU MIXED1 NORMAL
EXP14 HUAWEIP8 ANDROID6 EU MIXED1 FULL
EXP15 HUAWEIP8 ANDROID6 EU MIXED2 NORMAL
EXP16 HUAWEIP8 ANDROID6 EU MIXED2 FULL
EXP17 HUAWEIP8 ANDROID6 EU MIXED3 NORMAL
EXP18 HUAWEIP8 ANDROID6 EU MIXED3 FULL
EXP19 HUAWEIP8 ANDROID6 CHINA MIXED1 NORMAL
EXP20 HUAWEIP8 ANDROID6 CHINA MIXED1 FULL
EXP21 HUAWEIP8 ANDROID6 CHINA MIXED2 NORMAL
EXP22 HUAWEIP8 ANDROID6 CHINA MIXED2 FULL
EXP23 HUAWEIP8 ANDROID6 CHINA MIXED3 NORMAL
EXP24 HUAWEIP8 ANDROID6 CHINA MIXED3 FULL
EXP25 HTCONEM9 ANDROID6 EU MIXED1 NORMAL
EXP26 HTCONEM9 ANDROID6 EU MIXED1 FULL
EXP27 HTCONEM9 ANDROID6 EU MIXED2 NORMAL
EXP28 HTCONEM9 ANDROID6 EU MIXED2 FULL
EXP29 HTCONEM9 ANDROID6 EU MIXED3 NORMAL
EXP30 HTCONEM9 ANDROID6 EU MIXED3 FULL
EXP31 HTCONEM9 ANDROID6 CHINA MIXED1 NORMAL
EXP32 HTCONEM9 ANDROID6 CHINA MIXED1 FULL
EXP33 HTCONEM9 ANDROID6 CHINA MIXED2 NORMAL
EXP34 HTCONEM9 ANDROID6 CHINA MIXED2 FULL
EXP35 HTCONEM9 ANDROID6 CHINA MIXED3 NORMAL
EXP36 HTCONEM9 ANDROID6 CHINA MIXED3 FULL
EXP37 LGNEXUS ANDROID6 EU MIXED1 NORMAL
EXP38 LGNEXUS ANDROID6 EU MIXED1 FULL
EXP39 LGNEXUS ANDROID6 EU MIXED2 NORMAL
EXP40 LGNEXUS ANDROID6 EU MIXED2 FULL
EXP41 LGNEXUS ANDROID6 EU MIXED3 NORMAL
EXP42 LGNEXUS ANDROID6 EU MIXED3 FULL
EXP43 LGNEXUS ANDROID6 CHINA MIXED1 NORMAL
EXP44 LGNEXUS ANDROID6 CHINA MIXED1 FULL
EXP45 LGNEXUS ANDROID6 CHINA MIXED2 NORMAL
EXP46 LGNEXUS ANDROID6 CHINA MIXED2 FULL
EXP47 LGNEXUS ANDROID6 CHINA MIXED3 NORMAL
EXP48 LGNEXUS ANDROID6 CHINA MIXED3 FULL
EXP49 SAMSUNGS6EDGE ANDROID6 EU MIXED1 NORMAL EXP50 SAMSUNGS6EDGE ANDROID6 EU MIXED1 FULL
Table 4.2: Experimental plan of the case study – continued from previous page
ID DEV VER APP EVENTS STO
EXP51 SAMSUNGS6EDGE ANDROID6 EU MIXED2 NORMAL EXP52 SAMSUNGS6EDGE ANDROID6 EU MIXED2 FULL EXP53 SAMSUNGS6EDGE ANDROID6 EU MIXED3 NORMAL EXP54 SAMSUNGS6EDGE ANDROID6 EU MIXED3 FULL EXP55 SAMSUNGS6EDGE ANDROID6 CHINA MIXED1 NORMAL EXP56 SAMSUNGS6EDGE ANDROID6 CHINA MIXED1 FULL EXP57 SAMSUNGS6EDGE ANDROID6 CHINA MIXED2 NORMAL EXP58 SAMSUNGS6EDGE ANDROID6 CHINA MIXED2 FULL EXP59 SAMSUNGS6EDGE ANDROID6 CHINA MIXED3 NORMAL EXP60 SAMSUNGS6EDGE ANDROID6 CHINA MIXED3 FULL EXP61 SAMSUNGS6EDGE ANDROID7 EU MIXED1 NORMAL EXP62 SAMSUNGS6EDGE ANDROID7 EU MIXED1 FULL EXP63 SAMSUNGS6EDGE ANDROID7 EU MIXED2 NORMAL EXP64 SAMSUNGS6EDGE ANDROID7 EU MIXED2 FULL EXP65 SAMSUNGS6EDGE ANDROID7 EU MIXED3 NORMAL EXP66 SAMSUNGS6EDGE ANDROID7 EU MIXED3 FULL EXP67 SAMSUNGS6EDGE ANDROID7 CHINA MIXED1 NORMAL EXP68 SAMSUNGS6EDGE ANDROID7 CHINA MIXED1 FULL EXP69 SAMSUNGS6EDGE ANDROID7 CHINA MIXED2 NORMAL EXP70 SAMSUNGS6EDGE ANDROID7 CHINA MIXED2 FULL EXP71 SAMSUNGS6EDGE ANDROID7 CHINA MIXED3 NORMAL EXP72 SAMSUNGS6EDGE ANDROID7 CHINA MIXED3 FULL
The first set (EXP13∼EXP60) covers all of the DEV levels, and keeps the VER factor to ANDROID6, since Android 6 Marshmallow is the only version that can be installed on all the devices, allowing us to study the impact of software aging across devices from different vendors (and all other factors with the same level). The second set of experiments (EXP1∼EXP24) fixes the DEV factor to HUAWEIP8, and varies the VER factor between ANDROID5 and ANDROID6. The third set (EXP49∼EXP72), instead, fixes DEV to SAMSUNGS6EDGE and the VER to either ANDROID6 or ANDROID7. These last two sets are used to study the impact of software aging across different versions of the Android OS. In each set, with the sole exception of the fixed factor, we consider every possible combination of the levels, leading to a full factorial design. Based on our preliminary experiments [60],
USB CABLE
Workstation
Database
Commands: monkey, logcat Test Configuration: app, storage
Activity Launch Time PSS Measurements GC Measurements Task-level Measurements
FIGURE4.1: The Experimental Android Testbed
we calibrated the duration of each experiment to 6 hours, as this duration suffices to point out software aging effects.
The devices are controlled and monitored using the Android Debug Bridge (ADB) utility (which is a non-intrusive, dedicated channel through the USB port for debugging purposes). User inputs are provided with the monkey tool, which is a workload generator that randomly generates UI events. The events are generated at a high frequency (500ms) to stress the device, and follow the random profile of the EVENTS factor. The experimental testbed is showed in Figure4.1.