Task offloading application often need to offload the task at a particular point of time. Thus, we need to find an optimal computational node to offload the task at that time. Estimating the usage statistics of the available computational nodes at a particular time in near future can help to determine the optimal computational node.
Every 15 minutes, this app logs following paramters with a timestamp,
- available batery%,
- RAM usage percentage
- CPU frequency
- available internal storage (in KB) )
The values are written to a .csv file with the aim to create a dataset. This dataset will be used to predict the system paramters
of the phone at a particular time. Even if the user switches off the phone (or battery runs out) the app will start logging again when
the phone is turned on again. The user can explicitly stop background logging from the app.
Google Work Manager API has been used to handle the periodic background work of logging.
This app does not do any network activity which means the app does not send the data to any server. The log file is stored on the external emulated storage of the app folder. It makes the log file visible to the user so that the user can be confident about the data that is being logged.
No indivduality or unique identifier of the device or of the user is stored in the log file or within the app.
The logging work runs in background every 15 minutes. Since, different OEMs implement their own battery optimizers, they tend to aggressively kill the background tasks. Click here to know more on this. As a workaround, the app should be exempted from battery restrictions manually.
Currently, the future prediction of system paramters is being done in separate project on an experimental basis. We can try doing this prediction within the app itself.