2016 International Conference on Service Science, Technology and Engineering (SSTE 2016) ISBN: 978-1-60595-351-9 Research on the Checkpoint Server Selection Strategy Based on the Mobile Prediction in Autonomous Vehicular Cloud Wen-Di CHI a, Ru LI b,*, Peng-Fei FAN c Inner Mongolia University Computer College, China a chiwendi@126.com, b liru@imu.edu.cn, c lyzfpf@163.com *Corresponding author Keywords: Autonomous Vehicular Cloud, Mobile Prediction, Checkpoint Server, OPNET. Abstract. With the continuous development of wireless communication technology and the improvement in the performance of the vehicular device, the concept of Autonomous Vehicular Cloud (AVC) has emerged. Vehicular Cloud can provide abundant resources and services, but the high mobility of highway vehicle nodes leads to the network communication time is short, therefore allocated resources lost easily. This paper will focus on solving the problem of the low efficiency of the resource aggregation and distribution of the vehicle nodes in the highway scene. Based on the particularity of the Vehicular cloud in the highway, this paper introduces the checkpoint server selection mechanism which is based on mobility prediction and proposes two checkpoint server selection schemes, one is the choice of relative task initiating node stability of the vehicle node, the other is the choice of relative task executing node stability of the vehicle node. This paper uses OPNET Network simulation tool to evaluate the two schemes mentioned above based on various scenarios of different node density. Finally, the empirical conclusion is that the selected checkpoint based on mobility prediction methods is more stable, can improve the efficiency of resource aggregation and allocation, and vehicle node is more sparse, the enhanced effect is more obvious. Introduction As the improvement of the performance of modern mobile communication devices, the Vehicular Ad hoc Network (VANET) has attracted much attention. On the other hand, cloud computing it still steady development with its irreplaceable position in recent years. The background that this paper based on is combining VANET with cloud computing organically, named Autonomous Vehicular Cloud (AVC) [1]. In AVC, vehicle nodes has the following characteristics, for the urban road, vehicle nodes are prone to accumulate, leading to a dense distribution and a low average speed of vehicle nodes; And for the highway, vehicle nodes distribute non-uniformly, the average speed of vehicle nodes is high, the phenomenon of overtaking among vehicle nodes is frequent. In order to solve the problem mentioned above, this paper focuses on improving the efficiency in resource aggregation and distribution of vehicle nodes in the highway. In addition, this paper introduces the checkpoint server, and proposes the checkpoint server selection mechanism based on mobility prediction to improve the the efficiency of resource aggregation and allocation. In this paper, we introduce the checkpoint server should be described in this way: Firstly, a backup server, saves part of resources had been allocated or collected; secondly, a transfer station, it can provide quickly and effectively, when this part of the resources is required by other nodes. 262
Related work Olariu in 2010 proposed the concept of AVC [2], in recent years related studies are also the discussion of the concept of Vehicular Cloud, application scenarios and framework [3,4]. The studies relate to the mechanism of AVC resources aggregation and allocation are basic in the blank, however the resources Vehicular Cloud can provide are huge. Therefore, how to solve resource aggregation and allocation efficiently is a crucial problem. In the literature [5], the authors propose a coordination point selection mechanism, and implement local clustering dynamically based on the coordination point selection protocol, so the coordination point can aggregate resources clustering. The author uses the cluster structure in the network strong communication ability to achieve the resources aggregation and distribution, is to build clusters and maintenance more efficient, the more the efficiency of resource aggregation and distribution. However, the network environment proposed by the coordination point selection mechanism in the literature [5] is also in AVC, but their research objects are the vehicles in the city road. Because the distribution of vehicles in the city road is dense, travel speed is slow, so this coordination point selection mechanism is very suitable, and this is also the reason that the survival time as an important basis for the choice. While facing the highway vehicle nodes, the speed is faster than urban road nodes, overtaking phenomenon is extreme common, these reasons make the coordination point selection mechanism is not suitable for this paper. Also, this coordination point selection mechanism manages resources within the clusters by cluster structure, and the particularity of the highway vehicle nodes show that to build clusters among highway vehicle nodes is an unwise choice. Because it is not suitable to build cluster in the highway, therefore in literature [6],making cluster head node as the guideline for choosing checkpoint server is not workable in this paper, however, authors mentioned choosing the neighbor node of the task undertaking node as the checkpoint server in the literature [7], broadening the mind for this paper to put forward to the checkpoint server selection mechanism based on mobility prediction. In literature [8], it judges the link stability by calculating the Link Expiration Time (LET), this metod is mainly used to improve the efficiency of routing protocol. According to the method of judging the link stability mentioned above, this paper puts forward to the method that using the vehicle node mobile information to estimate Link Expiration Time between the vehicle nodes. Checkpoint Server Based on Mobile Prediction Mobile Prediction Model Descriptions The mobile prediction model in this paper is applicable to communicate with the vehicles in highway without the base station and the requirements of the design is the fast method to calculate the link stability, specifically, using the mobile information of vehicle nodes, calculating the Link Expiration Time between the nodes hop by hop, and according to the Link Expiration Time of the bottleneck link to judge the stability of the whole link, so as to choose a relatively stable link. 1) The stability of the link 263
Due to the high mobility of the vehicle nodes in VANET, making the topology of the network change more frequently than other mobile Ad hoc network, the way to find a lasting, stable and efficient communication path in VANET whose topology changes frequently has been the focus of attention in academia. The stability prediction method of the communication path selection of communication path improves the efficiency of communication at a certain. Usually one communication path is composed of multiple links, and the stability of the whole path depends on the stability of the bottleneck link. 2)Calculation of Link Expiration Time The basic formula, as shown in the formula 3.1: LET = R + V α * i ± V D j ij (1) LET: the Link Expiration Time between nodes R: the longest transmission distance of one hop D ij : the distance between two nodes V i, V j : the speed of the node α: as a parameter, the value is±1. When V i < V j, the value of α is 1; When V i > V j, the value of α is -1. Assuming at a certain moment, two nodes i and j can communicate with each other directly, as shown in Fig.1: Figure 1. The Calculation Method of Link Duration Schematic of First Situation. All of the above mentioned conditions are the same direction of the two node, if the direction of the two nodes is opposite, only the relative speed will be V i +V j, while the other parts are unchanged. The Process of the Checkpoint Server Selection based on Mobility Prediction Selecting the checkpoint server is divided into several stages: 1) The task initiating node send hop inquiry packet, which contains the location information, speed information of the current node, and IP address of task initiating node; 2) When the normal node receives the hop inquiry packet, it has the following actions: First, recording the IP address of the forwarding node and getting the data in the packet hops; Second, calculating the LET between the current node and the last hop node; Third, filling the hop inquiry packet with the current nodes speed and position information, and forwarding the hop inquiry packet. Finally, filling the hops reply packet with the information of the hop between the current node and the task initiating node, the IP address of the last hop node, and the LET between the current node and the last hop node, and forwarding the hop reply packet; 264
3) After the task initiating node receives all the hop reply packets, it has the following actions: First, recording the data in the hop reply packet, and selecting a task execution node randomly among the nodes which have replied the hop reply packet; Second, according to the last hop IP address of all the hop reply packets can construct the global topology between the task initiating node and task executing node; 4) Comparing the LET of the bottleneck link in each link, and the link whose LET of the bottleneck link is the longest of the relatively stable link; 5) According to the selected stability link, finding the bottleneck link and selecting the checkpoint server. OPNET Simulation Environment and the Analysis of the Experimental 1) Task instance description This paper in the experimental stage, assigns the task of large number factorial for vehicle nodes, participating with many nodes, distributed execution, followed by the no checkpoint server, selecting the checkpoint server based on mobility prediction (the choice of relative task initiating node stability, the choice of relative task executing node stability), selecting the checkpoint server randomly, all kinds of schemes are carried out the simulation respectively in different network topologies. In order to make the task instance have little effect on the final experimental results. This paper choose to the factorial of 1000 as the task instance, divide the task into multiple steps and assign to the task executing node sequentially. This task has such benefits, it no depends on execution order between sub tasks, and the calculation of the factorial of 1000 compared with general simple task is time-consuming. In a certain extent, this task can represent a class of task. This paper sets up six scenarios, each of which is independent of each other, and the number of nodes in each scene is 50, 60, 70, 80, 90, 100.Six scenarios and four kinds of comparing scheme, according to the principle of controlling variables, it needs to carry out 24 groups of experiments. In order to compare the advantages or disadvantages of different checkpoint server selection scheme, the number of simulation times is 1000, each simulation only assigned one task, recording the task execution is successful or not after each simulation, finally, counting the rate of task completion of 1000 times. 2) Introduction of simulation environment The specific settings of OPNET simulation parameters, as is shown in Table.1: Table 1. The Simulation Parameters of OPNET. Parameter Type Simulation Platform Parameter Value Windows7+OPNET14.5 Number of nodes 50,60,70,80,90,100 Wireless Transmission Protocol 802.11b Transmission Data Rate 11Mbps Radio Signal Transmission Power 0.005W Packet Received Power Threshold -76dBm Signal Propagation Distance 125m Routing Protocol AODV 265
The original intention of using the VanetMobiSim traffic simulation tool is to establish a scene as close as possible to the real highway. Therefore, it is very important to construct a suitable traffic flow scene. According to the needs of users, VanetMobiSim can configure the corresponding parameters in the XML file to generate the vehicle trajectory file, the specific parameters settings of VanetMobiSim, as is shown in Table.2: Table 2. The Simulation Parameters of VanetMobiSim. Parameter Type Parameter Value Range m 2 5000*200m 2 Initial Position Random Minimum Speed m/s 19.4m/s Maximum Speed m/s 33.3m/s Maximum Acceleration m/s 2 0.6m/s 2 Maximum Deceleration m/s 2 0.9m/s 2 Safe Distance m 100m Vehicle Lane Two-way 4 Lane Mobility Model IDM_LC 1) Simulation results In this section, the simulation results of four kinds of checkpoint server selection schemes in six scenarios will be given. The abscissa of the simulation results marks the number of vehicle nodes, the ordinate marks the rate of task completion of 1000 times, the simulation results is shown in Fig.2: Figure 2. Simulation Result. 2) Conclusions of the experiment After a comprehensive comparison of the simulation results, the conclusion is that compared to not choosing the checkpoint server and selecting the checkpoint server randomly, the method based on mobility prediction can make the task completion rate higher. It is enough to explain selecting the checkpoint server based on mobility prediction method is reasonable. Also based 266
on mobility prediction method, compared the relative task initiating node stability with the relative task executing stability node, choosing the former can make the task completion rate higher; In addition, the improve of task completion rate is the most obvious in the situation that vehicle density is smaller. This conclusion is reasonable, because with the increase of the number of vehicle nodes, the whole network topology tends to be stable. Therefore, when the vehicle density is high, task completion rate tends to be stable. Remarks This paper proposes the checkpoint server selection scheme based on mobile prediction in AVC, and builds a stable link by calculating the LET between nodes. Then this paper compares the different checkpoint server selection scheme, and proves the advantage of the method based on mobility prediction, finally draw the conclusion that choosing the relative task initiating node stability of vehicle node as the checkpoint server is more reasonable. Acknowledgement This research was financially supported by The National Natural Science Foundation. References [1] Eltoweissy M, Olariu S, Younis M. Towards autonomous vehicular clouds [M]. Ad hoc networks. Springer Berlin Heidelberg, 2010: 1-16. [2] Lochert C, Scheuermann B, Wewetzer C, et al. Data aggregation and roadside unit placement for a vanet traffic information system[c]. Proceedings of the fifth ACM international workshop on VehiculAr Inter-NETworking. ACM, 2008: 58-65. [3] Olariu S, Hristov T, Yan G. The next paradigm shift: from vehicular networks to vehicular clouds [J]. Basagni, S. and Conti, M. and Giordano, S. Stojmenovic, I), (Eds), Mobile Ad hoc networking: the cutting edge directions, Wiley and Sons, New York, 2012. [4] Olariu S, Khalil I, Abuelela M. Taking VANET to the clouds [J]. International Journal of Pervasive Computing and Communications, 2011, 7(1): 7-21. [5] Jingguo Ma. A Research on the Cluster Coordinator Selected Mechanism based on Autonomous Vehicular Cloud [D]. Inner Mongolia University, 2014. [6] T.Y.T. Juang and M. C. Liu, An EfficientAsynchronous Recovery Algorithm In Wireless MobileAd HocNetworks, Journal of Internet Technology, Vol. 3, No.2, 2002, 143-152. [7] A. K.Singh, P. K. Jaggi, Staggered Checkpointingand Recovery in Cluster Based Mobile Ad HocNetworks,International Conference on Parallel,Distributed Computing technologies andapplications Springer Proceedings 2011. [8] Namboodiri V, Gao L. Prediction-based routing for vehicular ad hoc networks [J]. Vehicular Technology, IEEE Transactions on, 2007, 56(4): 2332-2345. 267