Dynamic Code/Data Offloading in Cloud Computing Architecture

According to Forbes, over 80% of entire enterprise workloads will be in the cloud by 2020. Cloud computing is of utmost importance in future 5G+/6G network systems. We aim at developing essential technologies in cloud computing network architecture: code and data offloading policies using computing and networking resources. The policies can be developed by various optimization techniques and AI-based algorithms (e.g., deep Learning and reinforcement learning). Moreover, the developed algorithms are evaluated by simulations (e.g., using MATLAB) or experiments (e.g., using smartphones and Amazon Web Service (AWS)).

Reference Papers

  • Collaboration of Network Providers and Cloud Providers, IEEE Network, Sep. 2020
  • DREAM: Dynamic Resource and Task Allocation in Mobile Cloud Systems, IEEE JSAC, Dec. 2015
  • Dual-side Code Offloading System, IEEE TVT, Feb. 2018

  • Elastic Content Caching in Edge Computing Architecture

    According to Cisco, mobile video traffic will account for more than 80% of total mobile traffic by 2022. These traffic can be significantly alleviated by proper edge caching at world-widely distributed cloud servers. Content Providers (CPs) without CDN servers can rent cache space as needed at different cloud locations from cloud service providers (CSP) such as AWS in order to enhance their offered quality of service (QoS). We addresses key challenges in this context, namely how to invest an available budget in cache space to match spatio-temporal fluctuations of contents demand, wireless environment and storage price. To estimate popularity of each content file, we exploit not only big data and machine learning paradigm but also optimization techniques. Content caching system can be developed by physical servers in the lab and virtual servers rented by cloud service providers, e.g., AWS.

    Reference Papers

  • ElastiCache: Elastic Budget Sharing in Content Caching System, IEEE TWC, 2021
  • Hierarchical Content Caching System, IEEE TWC, May 2018
  • Joint Content Caching and Routing, IEEE TCOM, Aug. 2019

  • Mutli-resource Management for Real-time AI in IoT networks

    Recent AI applications such as Caffe2Go of Facebook or face recognition algorithm in VR/AR services require semi-real time processing. Hence, most of works must be processed at the mobile or IoT devices or edge servers close to the end users to reduce end-to-end latency of the mobile services. However, compute/storage/network/battery resources of the compact-sized devices are extremely limited. Therefore, we aim at developing multi-resource (compute/storage/network/battery) management policies by sharing resources among proxy IoT/mobile devices to maximize both efficiencies of energy consumption and service latency. To evaluate the algorithms, we exploit Android/iOS smartphones and AI-viable IoT devices.

    Reference Papers

  • Information Sharing Architecture for Mobile Devices, IEEE TMC, Feb. 2019
  • Multi-resource Management in a Smartphone, IEEE ToN, June 2016

  • Radio Resource Management for 5G Communications

    5G communications support various types of applications which ask different network bandwidth and latency requirements. We aim at developing key radio resource management techniques in 5G communications such as network slicing, beam activation, power control and user scheduling in CoMP (Coordinated MultiPoint) environment. Robust and convex optimization tools are used to design these techniques.

    Reference Papers

  • Virtual beamforming and user scheduling in mmWave network, IEEE COMML, Jan. 2019
  • Robust Power Allocation in Cognitive Radio Network, IEEE WCL, Oct. 2016
  • Optimization for Network Slicing, IEEE PIMRC, Oct. 2017