The Research On Key Technology Of Efficiency On Mobile Cloud And Deployment

With the smart phone taking the place of traditional cell phone, the computingcapability of mobile device is getting stronger. At the same time more and moresensors are installed on the devices to satisfy the demand of users and enhance users’experience. But the shortage of power formed a new problem which is focused bypeople. As for the mobile device, user wish not to use the functions to save moreenergy. So even if there’s strong features, they mean nothing. At the same time, as anew form of commodity, cloud computing comes into commercial. With thedevelopment of public and private cloud, datacenters are built in every city. Then asurprising idea that use cloud computing to solve the lack energy problem andincreasing computng speed of mobile devies was proposed. Since cloud service canprovide more resources such as saving storage and computing resource for mobiledevice, which means more energy in the mobile device can be used on moremeaningful tasks such as collecting data from sensors. For now, the research amongdomestic and foreign focus on realizing such a system, which can analyse programand measure the cost of each part of it by offline. Then they will devide the programinto parts and choose one sub of them to offload to the could server. But the pre-processing work is complicated and it’s hard to do such work for every program, andthe work of measuing cost seems impossible for developers to handle.The paper propose an online analysis program method. When a program is runon the mobile device, it will construct a method invoking tree which can reflect thestructure of the program. During the construction, we also do statistics for eachparameters of the tree. Then we propose a problem which takes the maximum thesaving time of running program as the optimization objective based on the methodinvokeing tree, and analysis it is not an NP-Complete problem. After that we give andynamic programing algorithm to solve such a problem, and prove it is true. And weanalysis the assumption of pre-synchronization of application file among thedomestic and foreign research next, and we think about a method to devide theprogram file. Just send the necessary part of codes and datas which will be run oncloud to the cloud server, and reserve the remain parts. In such way, we can save thenetwork resource and server disk. At last, we realize the features of construction method invoking tree online, offloadmethods set computing online and migrating program code by necessary. Wedesigned a group of experments to measure the speed of mobile devices and cloudserver. Then three different groups of experiments which satisfies3differentdemands for develops are conducted to test the function of construction methodinvoking tree and offload methods set computing, including matrix multiplication,searching prime by sieve and file processing. The experiments show that the speed-up are7.11x,23.23x,8.24x. At last we do the experiment for migrating code bynecessary, the experiment shows that the code migration by necessary can savenetwork by77.1,73.5%,76.2%, and save the disk by81.9%,77.8%,80.1%.