Latency Optimization On Computation Offloading Of Mobile Terminals

Mobile terminal,such as intelligent mobile phone,tablet,wearable device and net-work camera,is playing an important role with an explosive growth in daily life.While its development is faced with two challenges in terms of computation and communica-tion.On one hand,in order to support mobility,resources on mobile terminals such as CPU,GPU,storage capacity,memory,and battery lifetime are usually limited.Thus the resource-constrained mobile terminals can hardly meet the requirements of computation intensive applications such as big data processing,which is regarded as contradiction between mobility and processing ability from perspective of computation.To mitigate such a problem,computation offloading provides a solution where mobile terminals of-fload computation to the resourceful big-nodes on the transmission path.Mobile cloud computing(MCC)and mobile edge computing(MEC)are two technologies for provid-ing the big-nodes in computation offloading scenarios.However,offloading tasks to the cloud unavoidably introduces a high transmission delay,which cannot adapt to the delay-sensitive applications such as virtual reality(VR)and cloud gaming.This is regarded as contradiction between real-time requirement and service latency from perspective of communication.Under these circumstances,as the main challenge in latency optimization on computation offloading of mobile terminals,communication and computation coop-eration(3C)not only means meeting the computation requirements of mobile terminals,but also minimizing the communication latency.To optimize the computation offloading latency,we take 3C as the basic methodol-ogy,both MCC and MEC as the case studies,and Matching Theory as the optimization method.In this case,three aspects including the basic framework of latency optimization in computation offloading,the latency optimization in MCC-based offloading,and the effi-cient MEC-based offloading control are studied.The main contributions are summarized as below.1.Basic framework of latency optimization in computation offloading.With regard to the contradiction between communication and computation,we bring about a 3C-based duplex matching framework by applying the Matching Theory on big-node selections of mobile terminals,which is the basic framework for both the MCC-based and MEC-based latency optimization.Experimental results show that 3C-based solution can be used to reduce over 100%latency.2.Latency optimization in MCC-based offloading.To reduce the latency of the cloud distributed interactive applications(CDIAs),we conduct measurements on the real systems,and put forward the concept of virtualization latency by modeling the interferences between computation intensive tasks and bandwidth intensive tasks.Then we study the design framework of CDIA systems and propose the duplex matching scheme between cloud proxies and clients under the constraints of resource allocation on proxies.Experiments show that with the provisioning of real systems,our solution not only smartly allocates workloads but also dynamically assigns capacities across proxies based on their arrival/departure patterns.3.Efficient MEC-based offloading control framework and deployment of an MEC testbed.Considering the interferences in both communication and computation,we study the computation offloading in the hybrid architecture of cloud radio access network(C-RAN)with MEC and give a latency-sensitive offloading control framework.By applying the 3C-based duplex matching framework,we propose a feedback-based iterative algorithm to handle the multi-stage matching among users,antennas,base stations,and cloud virtual machines.Furthermore,an LTE-based MEC testbed is implemented,and trace-based simulation shows that our scheme can achieve near-optimal performance in 90%cases.