READ ME File For 'Dataset for " QUAREM: Maximising QoE through Adaptive Resource Management in Mobile MPSoC Platforms "'


Dataset DOI:  https://doi.org/10.5258/SOTON/D2153


ReadMe Author: Samuel Isuwa, University of Southampton [ORCID: 0000-0002-2235-4091] [email: s.isuwa@soton.ac.uk]


This dataset supports the publication:

AUTHORS: Samuel Isuwa, Somdip Dey, Andre P. Ortega, Amit Kumar Singh, Bashir M. Al-Hashimi, Geoff V. Merrett

TITLE: QUAREM: Maximising QoE through Adaptive Resource Management in Mobile MPSoC Platforms

JOURNAL: ACM Transactions on Embedded Computing Systems


This dataset contains:

'fig1.csv': Data supporting Fig. 1. Average daily plug-in status of approximately 50 smartphone users over a period of one year. A plug-in status of 1.0 means that all 50 users had their smartphones plugged in, whereas a plug-in status of 0.0 means that all were unplugged.

'fig2.csv': Data supporting Fig. 2. Different energy usage patterns and plug-in (charging) times experience by a single user.

'fig5.csv': Data supporting Fig. 5. Janks (dropped frames) observed for Google Maps while varying the frequency of the (a) GPU (b) big CPU and the (c) little CPU cores.

'fig6.csv': Data supporting Fig. 6. Janks (%) and mean operating frequencies (GHz) observed for YouTube, Google Maps and Chrome workloads while capping the frequency of the CPU.

'fig7.csv': Data supporting Fig. 7. Normalised energy consumption for Google Maps and Chrome while capping the frequency of the big CPU cores.

'fig8.csv': Data supporting Fig. 8. CPU frequency traces of the different approaches while varying Google Maps application.

'fig9.csv': Data supporting Fig. 9. Comparison between predicted plug-in time vs the Actual plug-in time of the user.

'fig10.csv': Data supporting Fig. 10. Comparison between the predicted energy demand vs the actual energy demand of the user across the time steps for a particular day. The energy demand plot is normalised as a percentage of the battery level.

'fig11.csv': Data supporting Fig. 11. Battery discharge profile for the different approaches considered.

'fig12.csv': Data supporting Fig. 12. Summary of the percentage of the final battery level and the percentage of time each approach was active at different initial battery level.

'fig13.csv': Data supporting Fig. 13. Comparison of instantaneous QoE for five different resource management techniques (higher is better) while running Google Maps and YouTube repeatedly. The QoS is measured as frames rate jank at every second.

'fig14.csv': Data supporting Fig. 14. Comparison of average QoE for five different resource management techniques while starting the day at different initial battery level (%).

'fig15.csv': Data supporting Fig. 15. Comparison of average QoE for five different resource management techniques while considering erratic charging patterns and starting the day at different initial battery level (%).


Date of data collection: March 2020 - July 2021


Information about geographic location of data collection:  UK


Licence: CC BY 4.0


Related projects:
PTDF 1526/19


Date that the file was created: March, 2022