A Promising CUDA-Accelerated Vehicular Area Network Simulator Using NS-3 Chok M. Yip and Abu Asaduzzaman Department of Electrical Engineering and Computer Science Wichita State University, Wichita, Kansas, USA
[email protected] Abstract—Both size and computational activities of Vehicular Area Network (VANET) are growing. Simulation of VANETs not only requires the simulation of network standards, but also requires the mobility of nodes. Such a dynamic system involves computations of node distances, routing protocols, application layers, data send, data receive, etc. The simulation model of VANET requires both hardware and software supports to deal with massive computational problems. Currently available network simulators, like Network Simulator 3 (NS-3), are not adequate for simulating large-scale VANET systems. In this work, we propose a Compute Unified Device Architecture (CUDA)-assisted VANET simulation model for multicore Central Processing Unit (CPU) / manycore Graphics Processing Unit (GPU) platform to increase computational throughput. The proposed VANET/GPU simulator uses NS-3 as the core engine and improves throughput by exploiting massively parallel processing on the GPU. Experimental results show that the overall computation speedup can be increased up to 129x by using the proposed VANET/GPU simulator. Keywords-Computational throughput, CUDA architecture, GPU computing, NS-3, modeling and simulation, VANET
I. INTRODUCTION VANET is a subset extension of Mobile Area Network (MANET) [1]. Communication basis of VANET alters between Vehicle to Infrastructure (V2I), Vehicle to Road-side unit (V2R), and Vehicle to Vehicle (V2V) [2]. In such a dynamic environment, changes in network topology occur frequently as distance between nodes changes dynamically. In addition, urban environment such as buildings and tress will interfere with microwave propagation [3]. These factors need to be considered while engineering a reliable network. Mobility module simulates driver and environment behaviors such as road obstacles and distances. Wireless access in vehicular environments (WAVE) model simulates dynamic short-range connection and channel behaviors specified by IEEE 802.11p standards. The simulation of IEEE 802.11p standard is provided by PhySim module [4]. This model provides behavior of physical layer such as channel switching, wireless signal interference, and wireless signal collision. The data-link layer amendment of IEEE 802.11p standard is provided by NS-3 framework. Mobility modeling consists of points and coordinates. This provides a basis to compute distance between network nodes. Obstacles between network nodes need to be accounted. The models developed by Benioff et al [5] can be used for accuracy, which has been validated against Treiber Intelligent Driver model. Bidirectional simulation between vehicles and network may be achieved by This research is partially supported by Kansas NSF EPSCoR First Award (WSU Fund #R50975) project.
978-1-4799-7575-4/14/$31.00 ©2014 IEEE
using previously developed models [6]. Applications like obstacle warning alerts should change driver’s driving method. Simulation and modeling has given human the ability to gain perception and develop solutions for complex systems in a fast, safe, and cost effective manner. VANET simulator allows network designers to gain different perspectives in an environment before expensive equipment needs to be purchased. Such methods have been utilized by network designers before deploying equipment on network. NS-3 is proven to be an effective vehicular network simulator [7]. However, traditional NS-3 computations are very time intensive with current networking technology. According to the recent studies, CUDA-assisted GPU computing has potential to significantly decrease the overall execution time [8]. Therefore, we introduce a parallel computing technique using CUDA/GPU to accelerate NS-3 computational throughput by modifying the mobility module. Sad; II.
PROPOSED VANET SIMULATOR
As we know, NS-3 provides a base-framework of network simulation. NS-3 is an open source C++ and Python simulator; NS-3 modular design allows users to add, modify, or replace modules based on requirements. We modify NS-3 to accommodate CUDA/GPU parallel computing and use the modified NS-3 as the core of the proposed VANET/GPU simulator. Simulation parameters such as map size, number of nodes, obstacles, and driver behavior model are defined in this instance [5]. Figure 1 illustrates modules provided by NS-3 framework. Each module can be replaced by user’s program; e.g., by modifying mobility module, we can implement vehicular mobility with waypoint and manage memory for GPU. The detailed internal architecture is explained by Gustavo [9].
protocols
test helper propagation
internet
Etc. mobility
network core Figure 1: Internal NS-3 architecture This model is organized into two streams; each stream holds and generates output. Figure 2 represents a process flow of NS-3. Using separate streams of data allows data to be
updated independently; a non-synchronized model between CPU and GPU can be utilized to generate even higher throughput. Simulation Procedure GPU
III.
RESULTS AND DISCUSSION
According to the experimental results (as illustrated in Figure 5), proposed CPU/GPU-VANET simulator may achieve up to 129x and 75x speedup due to Kepler and Fermi GPU cards, respectively.
CPU
Highway & Obstacle
NS3 Core
Drivers Response
WAVE Channel behaviour OSPF Routing Table Environment Modeling
Process
Drivers Response
Figure 5: Speedup due to Kepler and Fermi over CPU
Data Capture
IV.
Figure 2: Execution model of NS-3 using GPU Highway module defines coordinates and node positions for each vehicle, which are crucial for calculation of distance. The simplified model of highway consists of north, east, south, and west junctions. Each node is mapped into Cartesian coordinate for computation of distance and possible vehicular routes. For each node in highway, physical distance is computed and delegated to physical layer processing module. The memory representation of a highway module is illustrated in Figure 3. The many-to-many mapping of a highway module can be easily accelerated using GPU. Each node can be computed using each CUDA thread, allowing processes to be parallelized. Vehicle 1
X position
Vector
In this work, we present a GPU-assisted VANET simulator using NS-3. We simulate a system with many (up to 55,000) nodes for many (up to 10,000) movements. Total execution times are obtained for four different implementations (Workstation without GPU, Workstation with Fermi GPU, Supercomputer without GPU, and Supercomputer with Kepler GPU). According to the experimental results, our proposed VANET/GPU simulator produces satisfactorily the same results as the traditional NS-3 does. For 55,000 nodes and 10,000 movements, the proposed VANET/GPU simulator achieves speedup about 129x using Kelper GPU. Speedup can be further exploited by integrating more parallel models into the NS-3 simulator.
Y position
REFERENCES [1]
Vehicle 2
X position
Vector
Y position
…. Vehicle N
[2]
X position
Vector
Y position
Figure 3: Memory representation of a highway module The distance between any two nodes is computed using the pseudo-code presented in Figure 4. Here, Px retrieves position of X and Py retrieves position of Y in Cartesian coordinate form. The distance is then computed using standard distance formula. This algorithm has a big-o notation of O(N2) when run on a CPU.
[3] [4]
[5]
[6]
[7]
[8]
Figure 4: Distance computation
CONCLUSION
[9]
P. Ranjan and K.K. Ahirwar, “Comparative study of vanet and manet routing protocols,” in Proceedings of the Int’l Conference on Advanced Computing and Communication Technologies (ACCT 2011), 2011. S. Olariu and M.C. Weigle, “Vehicular networks: from theory to practice,” in CRC Press, 2010. R.C. Johnson and H. Jasik, “Antenna engineering handbook,” in New York, McGraw-Hill Book Company, 1356, 1984. M. Segata and R.L. Cigno, “Simulation of 802.11 PHY/MAC: The quest for accuracy and efficiency,” in Conference on Wireless On-demand Network Systems and Services (WONS), pp. 99-106. IEEE, 2012. M.R. Benioff, E.D. Lazowska, et al, “Computational Science: Ensuring America's Competitiveness,” President's Information Technology Advisory Committee, June 2005, http://www.nitrd.gov/ Pitac/reports/20050609_computational/computational.pdf. F. Ricardo, P.M. d'Orey, and M. Ferreira, “Divert for realistic simulation of heterogeneous vehicular networks,” in IEEE Int’l Conference on Mobile Adhoc and Sensor Systems (MASS), pp. 721-726. IEEE, 2010. H. Arbabi and M.C. Weigle, “Highway Mobility and Vehicular Ad-Hoc Networks in NS-3,” in Proceedings of the Winter Simulation Conference, pp. 2991-3003, 2010. A. Asaduzzaman, C.M. Yip, and R. Asmatulu, “CUDA/C Based ‘Green’ Technology for Very Fast Analysis of Nanocomposite Properties,” in SAMPE Tech 2013 Conference, Wichita, Kansas, Oct. 21-24, 2013. G. Carneiro, “NS-3: Network Simulator 3,” in UTM Lab Meeting April, Vol. 20, 2010.
