Transmission Time Estimator for Social and Cloud Applications in Smartphones

Present day users tend to use Smartphone for heavy load multimedia file sharing, and require a lot of upload bandwidth. However, the bandwidth is an expensive affair and upload could take a lot of time in a limited bandwidth scenario. In this work, we aim to develop a solution to estimate time to share or upload multimedia content from a Smartphone to social networks or cloud servers. We propose machine learning approach for the estimation and use support vector regression (SVR) method. Our feature space for SVR includes network parameters such as available bandwidth and round trip time (RTT), received signal strength index (RSSI) as channel parameter, CPU load, Available Battery and file size as system parameters. Based on these features we estimate the time needed to upload the multimedia content. We evaluate our solution for accuracy using cross validation and our findings reveal, 84% of estimates are in 80% confidence interval for LTE and 80% in case of 3G. This estimate can help the user#x002F;system to make decision to schedule delay tolerant uploads in more favorable environment to maximize the Quality of Experience (QoE). We also show by simple experiment power savings of up to 40% when upload in bad signal is rescheduled to good signal conditions. We provide architecture of our solution in Android and implement it on Samsung J7 device.