Internet of Things. User-Centric IoT. First International Summit, IoT360 2014, Rome, Italy, October 27-28, 2014, Revised Selected Papers, Part I

Research Article

To Run or Not to Run: Predicting Resource Usage Pattern in a Smartphone

Download103 downloads
  • @INPROCEEDINGS{10.1007/978-3-319-19656-5_49,
        author={Arijit Mukherjee and Anupam Basu and Swarnava Dey and Pubali Datta and Himadri Paul},
        title={To Run or Not to Run: Predicting Resource Usage Pattern in a Smartphone},
        proceedings={Internet of Things. User-Centric IoT. First International Summit, IoT360 2014, Rome, Italy, October 27-28, 2014, Revised Selected Papers, Part I},
        proceedings_a={IOT360},
        year={2015},
        month={7},
        keywords={Smart phone Usage prediction Resource utilisation Machine learning Mobile cloud computing IoT Sensor data},
        doi={10.1007/978-3-319-19656-5_49}
    }
    
  • Arijit Mukherjee
    Anupam Basu
    Swarnava Dey
    Pubali Datta
    Himadri Paul
    Year: 2015
    To Run or Not to Run: Predicting Resource Usage Pattern in a Smartphone
    IOT360
    Springer
    DOI: 10.1007/978-3-319-19656-5_49
Arijit Mukherjee1,*, Anupam Basu2, Swarnava Dey1, Pubali Datta1, Himadri Paul1
  • 1: Tata Consultancy Services
  • 2: IIT Kharagpur
*Contact email: mukherjee.arijit@tcs.com

Abstract

Smart mobile phones are vital to the Mobile Cloud Computing (MCC) paradigm where compute jobs can be offloaded to the devices from the Cloud and vice-versa, or the devices can act as peers to collaboratively perform a task. Recent research in IoT context also points to the use of smartphones as sensor gateways highlighting the importance of data processing at the network edge. In either case, when a smart phone is used as a compute resource or a sensor gateway, the corresponding tasks must be executed in addition to the user’s normal activities on the device without affecting the user experience. In this paper, we propose a framework that can act as an enabler of such features by classifying the availability of system resources like CPU, memory, network usage based on applications running on an Android phone. We show that, such app-based classifications are user-specific and app usage varies with different handsets, leading to different classifications. We further show that irrespective of such variation in classification, distinct patterns exist for all users with available opportunity to schedule external tasks, without affecting user experience. Based on the next to-be-used applications, we output a predicted set of system resources. The resource levels along with handset architecture may be used to estimate worst case execution time for external jobs.