TY - GEN
T1 - Impact of device performance on mobile internet QoE
AU - Dasari, Mallesham
AU - Vargas, Santiago
AU - Bhattacharya, Arani
AU - Balasubramanian, Aruna
AU - Das, Samir R.
AU - Ferdman, Michael
N1 - Publisher Copyright:
© 2018 Association for Computing Machinery.
PY - 2018/10/31
Y1 - 2018/10/31
N2 - A large fraction of users in developing regions use relatively inexpensive, low-end smartphones. However, the impact of device capabilities on the performance of mobile Internet applications has not been explored. To bridge this gap, we study the QoE of three popular applications - Web browsing, video streaming, and video telephony - for different device parameters. Our results demonstrate that the performance of Web browsing is much more sensitive to low-end hardware than that of video applications, especially video streaming. This is because the video applications exploit specialized coprocessors/accelerators and thread-level parallelism on multi-core mobile devices. Even low-end devices are equipped with needed coprocessors and multiple cores. In contrast, Web browsing is largely influenced by clock frequency, but it uses no more than two cores. This makes the performance of Web browsing more vulnerable on low-end smartphones. Based on the lessons learned from studying video applications, we explore offloading Web computation to a coprocessor. Specifically, we explore the offloading of regular expression computation to a DSP coprocessor and show an improvement of 18% in page load time while saving energy by a factor of four.
AB - A large fraction of users in developing regions use relatively inexpensive, low-end smartphones. However, the impact of device capabilities on the performance of mobile Internet applications has not been explored. To bridge this gap, we study the QoE of three popular applications - Web browsing, video streaming, and video telephony - for different device parameters. Our results demonstrate that the performance of Web browsing is much more sensitive to low-end hardware than that of video applications, especially video streaming. This is because the video applications exploit specialized coprocessors/accelerators and thread-level parallelism on multi-core mobile devices. Even low-end devices are equipped with needed coprocessors and multiple cores. In contrast, Web browsing is largely influenced by clock frequency, but it uses no more than two cores. This makes the performance of Web browsing more vulnerable on low-end smartphones. Based on the lessons learned from studying video applications, we explore offloading Web computation to a coprocessor. Specifically, we explore the offloading of regular expression computation to a DSP coprocessor and show an improvement of 18% in page load time while saving energy by a factor of four.
KW - Hardware Accelerators
KW - Mobile Applications
KW - Quality of Experience
UR - https://www.scopus.com/pages/publications/85058160093
U2 - 10.1145/3278532.3278533
DO - 10.1145/3278532.3278533
M3 - Conference contribution
AN - SCOPUS:85058160093
T3 - Proceedings of the ACM SIGCOMM Internet Measurement Conference, IMC
SP - 1
EP - 7
BT - IMC 2018 - Proceedings of the Internet Measurement Conference
PB - Association for Computing Machinery
T2 - 2018 Internet Measurement Conference, IMC 2018
Y2 - 31 October 2018 through 2 November 2018
ER -