Research On Key Issues Of Scalable Real-time Online Interactive Application In Cloud

In recent years, real-time online interactive application (ROIA) is an emerging kind ofdistributed application based on Internet. The typical applications of ROIA include massivelymultiplayer online games and interactive e-learning system.In ROIA, the characteristics of user number dynamic changing and user operationuncertainty make greatly affected on the resource utilization and the ability of coping withunexpected load peaks. Therfore, the ROIA system usually should have a certain scalability toadapt to the dynamic changes, i. e. the ROIA system should have the ablity to adjust theoverall performance by changing the available hardware resources and scheduling ways. Agood scalability not only can improve the utilization of system resources, but also canenhance the ability of coping with unexpected load peaks. Thus the research on scalableROIA has important academic value and practical significance for promoting the developmentof ROIA.Currently in research area of ROIA, the most works are confined inside the scalabilityissues of a single ROIA without considering the complementary characteristics betweendifferent kinds of ROIAs. So there are some limitations on these research methods, and itlimits the improving of system resource utilization and the ability of coping with unexpectedload peaks.Aiming at this problem, the dissertation takes improving the resource utilization and theability of coping with unexpected load peaks in ROIA as main targets, and focuses on the keyissues of scalable ROIA in cloud. The key issues mainly include the fault-tolerant architectureof ROIA in the cloud, dynamic load balancing algorithm for multiple ROIAs, network latencyissue in ROIA and scalable architecture of multiple ROIAs in the cloud.The main research work and innovations of the dissertation include the following aspects:(1) To solve the problem which the peer failure is increasing greatly in the cloudenvironment and the traditional architecture of ROIA does not provide suitable fault tolerance,this dissertation proposes a scalable ROIA fault-tolerant architecture (SRFA). The SRFA addsa layer of fault-tolerant structure based on the general graph partitioning method. The virtual world of ROIA is covered by two layers of hexagonal cells, and the different servers areresponsible for the different regions consisted of cells in each layer. Using SRFA in ROIA willbe beneficial for seamless migration beteen regions, also beneficial for crash protection andsystem recovery, and also beneficial for achieving system load balancing and improving theutilization of hardware resources.(2) To solve the problem which it is difficult to improving the resource utilization and theability of coping with unexpected load peaks at the same time in a single ROIA, thisdissertation proposes a new scalable multi-ROIA dynamic load balancing (SMRDLB)algorithm. The SMRDLB algorithm takes advantage of some characteristics such as: differentrequirements on delay, different real-time interactive complexity, different times occurringload peaks. And the SMRDLB algorithm makes the multiple real-time online interactiveapplication system can acquire the higher resource utilization and the better ability of copingwith unexpected load peaks.(3) To overcome the increasing in network delay and network bandwidth demanding innew architecture, this dissertation proposes an improved dead reckoning algorithm based ongoal-predicting, called (goal-predicting based dead reckoning (GBDR). The GBDR algorithmimproves the traditional dead reckoning algoritym, and extrapolates the target at someappropriate time point in order to improve prediction accuracy. The improving of predictionaccuracy can reduce dead reckoning error, thereby reducing the sending of update informationand reducing the demand for network bandwidth.(4) This dissertation proposes a scalable multiple real-time online interactive applicationarchitecture in the cloud environment, called multi-ROIA cloud platform (MRCP). TheMRCP architecture takes full advantage of scale effect and complementary effect, and itmakes the multiple real-time online interactive application system have better scalability andcan adjust overall performance dynamically; in addition, the system resource utilization andthe ability of coping with unexpected load peaks have also been greatly improved.