Research Article  Open Access
Shugang Jiang, Yu Zhang, Zhongchao Lin, Xunwang Zhao, "An Optimized Parallel FDTD Topology for Challenging Electromagnetic Simulations on Supercomputers", International Journal of Antennas and Propagation, vol. 2015, Article ID 690510, 10 pages, 2015. https://doi.org/10.1155/2015/690510
An Optimized Parallel FDTD Topology for Challenging Electromagnetic Simulations on Supercomputers
Abstract
It may not be a challenge to run a FiniteDifference TimeDomain (FDTD) code for electromagnetic simulations on a supercomputer with more than 10 thousands of CPU cores; however, to make FDTD code work with the highest efficiency is a challenge. In this paper, the performance of parallel FDTD is optimized through MPI (message passing interface) virtual topology, based on which a communication model is established. The general rules of optimal topology are presented according to the model. The performance of the method is tested and analyzed on three high performance computing platforms with different architectures in China. Simulations including an airplane with a 700wavelength wingspan, and a complex microstrip antenna array with nearly 2000 elements are performed very efficiently using a maximum of 10240 CPU cores.
1. Introduction
The principle of FDTD is that the calculation region is discretized by Yee grid to make the components of and be distributed at time and space alternately [1]. Then, there are four (or ) components around each (or ) component. This character makes the algorithm parallel in nature and using this the Maxwell equation can be transferred to a set of difference equations. The electromagnetic fields can be solved at time axis step by step. Then the electromagnetic fields distribution in each time step later can be obtained by the original values and boundary condition [2].
Researches on MPI based parallel FDTD for simulating complicated models have been published in the past decade. In 2001, Volakis et al. presented a parallel FDTD algorithm using the MPI library, where they raised an MPI Cartesian 2D topology [3]. Andersson developed parallel FDTD with 3D MPI topology in the same year [4]. In 2005, the authors studied the optimum virtual topology for MPI based parallel conformal FDTD algorithm on PC clusters [5–7]. In 2008, Yu et al. tested the parallel efficiency of the parallel FDTD [8] on BlueGene/L Supercomputer successfully and gave a parallel efficiency of 4000 cores under the case of balanced loads.
Although there are many publications on parallel FDTD, few of them involve parallel FDTD simulations utilizing more than 10000 cores. Most of the papers focused on load balancing when parallel efficiency was concerned, in addition to a more precise rule to achieve the best performance that needs to be given, especially in the case of simulations using tens of thousands of CPU cores on supercomputers.
With these concerned, in this paper, the influence of different virtual topology schemes on parallel performance of FDTD is studied through a theory model analysis. Then some tests are made on National Supercomputer Center in Tianjin (NSCCTJ) and National Supercomputing Center in Shenzhen (NSCCSZ) to verify the feasibility of theory. With the proposed theory model, some electrically large problems whose parallel scales up to 10240 cores are provided in this paper. And the parallel efficiency is nearly 80% when 10240 cores of SSC were utilized for an array with nearly 2000 elements. To the best of our knowledge, the proposed method achieves one of the best efficiencies ever reached using more than 10 thousands of CPU cores.
2. Computation Resources from Supercomputers
The program is tested on different clusters in three supercomputer centers, National Supercomputer Center in Tianjin (NSCCTJ) [9], National Supercomputing Center in Shenzhen (NSCCSZ) [10], and Shanghai Supercomputer Center (SSC) [11]. The parameters of computation resources in this paper are listed in Table 1.

3. Communication Model for Parallel FDTD
Communication is the main factor which affects the parallel performance of parallel codes. Therefore, reducing the amount of communication in FDTD by adjusting the virtual topology is selected as the optimization target.
Assume that the communication time in one time step iswhere is communication delay time, is communication number, is transmission speed, and is the communication data amount of or . The calculated equation of each parameter is as follows:where , , and are the topology values in three directions and , , and are the grids number in , , and directions.
From (1), it is known that when the total communication data amount is the same, the different topology scheme may bring the different communication number , which comes to different total time .
Take Dawning 5000A as example, the parameters us~2.5 us and /(1.6563 Gb/s) [12]. Assume that the total grids are 1000 × 1000 × 1000 and total cores are 1000, now the total communication delay time (9.72 ms) is about less an order of magnitude than the total communication time (121 ms). Under this cores scale, the communication delay time is the secondary factor.
The communication amount of single process is
Divided by a constant , (4) will be
From (5), it is known when and only , namely, the topology conformal as calculation region, the communication amount of single process is the least. Generally, the equation above cannot be satisfied absolutely. So the topology distribution is required that it is divided along the direction which is conformal as calculation region as possible to make (5) the least.
Generally speaking, the communication time is less between processes in one node than that between processes which belong to different nodes [12, 13]; that is, the one byte data communication time factor is different between processes in one node and across nodes. So, when the factors and are the same between two different topologies, the communication amount across nodes is needed to be considered.
For certain grids, the total memory requirement (called ) keeps the same for different topologies. The memory distribution of each process (called ) is as follows:Equation (6) indicates that the memory distribution of each process is unrelated to the virtual topologies.
From the analysis above, it is known that the communication surface area varies from different virtual topology scheme for certain grids. The communication time will be changed associated with the virtual topology scheme while the memory distribution of each process remains the same. Thus, the communication amount is the main factor which affects the parallel performance.
4. Discussions on Parallel Performance
4.1. Simulation Model
Based on the theory above, a fourelement microstrip antenna array is used as the model for benchmark. The parallel FDTD code is run to analyze the virtual topology schemes on two supercomputer center platforms, National Supercomputer Center in Tianjin (NSCCTJ) and National Supercomputing Center in Shenzhen (NSCCSZ), as listed in Table 1.
The array model is shown in Figure 1. The parameters of this array are as follows: Central frequency is 4.97 GHz, mm, mm, mm, mm, , mm, mm, mm, and mm. The size of grid is mm.
Actually, the amount of total grids is just 200 × 200 × 50. However, to test the influence of different virtual topology schemes on parallel performance of parallel FDTD, the computational space needs to be extended. So, in this test, the amount of total grids is set as 1200 × 1200 × 300.
The radiation patterns of the microstrip array are shown in Figure 2, compared with the results obtained from HFSS. The figure shows that there is a well agreement between them.
(a) plane
(b) plane
4.2. Discussion of Parallel Performance
Here, we select several groups of virtual topology schemes to be tested. The following are the test results on the two supercomputer center platforms.
4.2.1. NSCCTJ
Table 2 is the comparison of total computation time (in seconds) with different CPU cores and different virtual topology schemes. The maximum number of CPU cores used for test is 120.

In Table 2, virtual topology schemes are described as () for all three communication patterns. If the value is 1 in some direction, it implies that there is no topology in this direction. For example, 2 × 1 × 1 means that there is no topology in and directions, respectively; thus the virtual topology is actually in one dimension. Similarly, 8 × 8 × 1 means that there is no topology in direction; thus the virtual topology is actually in two dimensions. In our work, one process uses one CPU core.
The speedup and parallel efficiency of the code are shown in Figure 3. From Figure 3, it can be seen that the parallel efficiency reaches up to 80% on NSCCTJ.
(a) Speedup
(b) Parallel efficiency
From Table 2, it is obvious that increasing the number of CPU cores can bring us the reduction of the computation time rapidly. But different virtual topology schemes will cost different computation time even if the code is run with the same number of processes. Next the parallel performance of the parallel FDTD will be discussed.
Here, the cases of 96 and 120 cores are taken as the examples. From (3)
The following is known.
(a) 96 Cores. Consider 8 × 6 × 2:(8 − 1) × (1200 × 300) + (6 − 1) × (1200 × 300) + (2 − 1) × (1200 × 1200) = 5760000 (946.52 s) (0.5 GB/process) 8 × 4 × 3:(8 − 1) × (1200 × 300) + (4 − 1) × (1200 × 300) + (3 − 1) × (1200 × 1200) = 6480000 (1500.54 s) (0.5 GB/process).
(b) 120 Cores. Consider 5 × 6 × 4:(5 − 1) × (1200 × 300) + (6 − 1) × (1200 × 300) + (4 − 1) × (1200 × 1200) = 7560000 (724.90 s) (0.4 GB/process) 6 × 5 × 4:(6 − 1) × (1200 × 300) + (5 − 1) × (1200 × 300) + (4 − 1) × (1200 × 1200) = 7560000 (1076.41 s) (0.4 GB/process).
From above, it is known that, for the virtual topologies with the same dimensions, the total grids at interfaces less (the amount of communication data less) can save calculation time effectively. Meanwhile, it is obvious that the memory distribution of each process is the same for different topologies with the same number of CPU cores.
But, from above, it also can be seen that, for the same amount of grids, the calculation time has certain differences. For example, the topologies 6 × 5 × 4 and 5 × 6 × 4 with the same grids number and the calculation time are 1076.41 seconds and 724.90 seconds, respectively. Generally, we believe that the consumption time between processes in one node is less and the more time consumed between ones across nodes [12, 13]. So, here we speculate that the different amount of the communication grids across nodes causes the difference between the two cases.
For 5 × 6 × 4, it has 4 × (1200 × 300) + 1 × (1200 × 300) = 1800000 FDTD grids needed to communicate across nodes, and its calculation time is 724.90 seconds, while, for 6 × 5 × 4, it has 5 × (1200 × 300) + (1 + 2/6) × (1200 × 300) = 2280000 FDTD grids needed to communicate across nodes, and its calculation time is 1076.41 seconds. The calculation way of communication data above is shown in Figure 4.
(a)
(b)
In Figure 4, every two adjacent figures with different colors have data communication. Figures 1 to 10 present ten nodes, and the adjacent figures with the same colors present that they are in the same node. For (b), the adjacent columns will transfer ((1 + 2/6) × (1200 × 300)) data in plane, and the adjacent rows will transfer (5 × (1200 × 300)) data in plane.
4.2.2. NSCCSZ
Table 3 is the comparison of total computation time (in seconds) with different CPU cores and different virtual topology schemes. The maximum number of CPU cores used for test is 480.

In Table 3, virtual topology schemes are described as () for all three communication patterns and the meaning of each figure in each topology is the same with the description in Section 4.2.1 above.
The speedup and parallel efficiency of the code are shown in Figure 5. From Figure 5, it can be seen that the parallel efficiency reaches up to 80% on NSCCSZ.
(a) Speedup
(b) Parallel efficiency
From Table 3, it can be seen that, for the virtual topologies with the same dimensions, the total grids at interfaces less (the amount of communication data less) can save calculation time effectively. This conclusion coincided with the case on NSCCTJ.
The amount of communication grids of each virtual topology is calculated by (3), while, for certain topologies with the same number of grids, it is found that the calculation time is less with the more crossingnode communication, analyzed by the way of NSCCTJ. It is contrary to the speculation theory above. Therefore, we speculate that whether it is caused by the reason that the communication amount of nodes with the main communication is great in the topology. Next, the cases of 10 × 2 × 3 and 3 × 5 × 4 in 60 cores are taken as the examples to analyze the speculation (the amount of communication is 6480000).
For 10 × 2 × 3, the crossingnode communication is 4 × (1200 × 300) + 1 × (1200 × 300) = 1800000, while, for 3 × 5 × 4, the one is 2 × (1200 × 300) + 16/12 × (1200 × 300) = 1200000. But the consumption time for 10 × 2 × 3 is 1117.17 seconds and the one for 3 × 5 × 4 is 1214.09 seconds. This does not agree with the speculation of crossingnode theory on NSCCTJ. To further explore the reason, the heaviest communication and the lightest communication of the related processes from these two virtual topology schemes are listed in Table 4. The heaviest communication load in the virtual topology scheme 10 × 2 × 3 is (1200 × 300)/30 + (1200 × 300) × 2/6 + (1200 × 1200) × 2/20 = 276000, while, in the virtual topology scheme 3 × 5 × 4, the one is (1200 × 300) × 2/12 + (1200 × 300) × 2/20 + (1200 × 1200) × 2/15 = 288000. The heaviest communication loads are assigned to the processes located at the center of the process grid. Similarly, the lightest communication loads that are set to the processes at the corner of the process grid can be calculated. The difference between the heaviest communication load and the lightest communication load in the virtual topology scheme 3 × 5 × 4 is larger than the one in 10 × 2 × 3, which results in different computation time. This indicates that when the total amount of communication is the same, a virtual topology scheme with a more balanced communication load may bring a better performance, although with a more crossingnode communication.

4.3. The General Rules of Optimal Virtual Topology
Generally, when the amount of FDTD cells is the same, the MPI virtual topology by the way where the amount of transferred data is less can save the computation time for the same dimensional virtual topology. The best performance of a parallel FDTD code can be obtained by optimizing the MPI virtual topology scheme. The general rules for a better performance are as follows.(a)Select MPI virtual topology scheme to make the total communication (equation (3)) the smallest.(b)When the total communication is the same, set the topology as with less crossingnode communication.(c)When the amount of crossingnode communication of different topologies is approximately the same, select the topology with a more balanced communication load.
5. Applications Using 10 Thousands of CPU Cores
Based on the optimal virtual topology scheme above, the parallel FDTD code is applied to analyze some complicated EM problems, and they are run on Shanghai Supercomputer Center (SSC) platform.
5.1. Radiation of Microstrip Antenna Array
5.1.1. Validation of the FDTD Code
To validate the FDTD code, a microstrip antenna array with hundreds of the same antenna units is analyzed and the results are compared with the ones provided by MoM. The model is shown in Figure 6. The amount of total grids is 786 × 1224 × 54. It is calculated on PC with the virtual topology scheme 4 × 3 × 2.
The radiation patterns of the array are shown in Figure 7, compared with the ones provided by MoM. From Figure 7, one can see that there are very well agreements between them.
(a) plane
(b) plane
5.1.2. Radiation of an Ellipse Antenna Array
An ellipse microstrip antenna array with nearly 2000 elements is shown in Figure 8. In this array, the amount of total grids is 6016 × 1160 × 54 (about 0.4 billion). It is tested on Shanghai Supercomputer Center (SSC), using 10240 cores, and the parallel efficiency from 4096 to 10240 cores is tested.
The parallel efficiency of this code is shown in Figure 9. And the radiation patterns of this array are shown in Figure 10.
(a) plane
(b) plane
From Figure 9, it is known that the parallel efficiency reaches to nearly 80% at 10240 cores with 4096 cores as benchmark, which achieves one of the best efficiencies ever reached using more than 10 thousands of CPU cores.
5.2. RCS of an Electrically Large Airplane
5.2.1. Simulation Model
The airplane is shown in Figure 11. The parameters are as follows: Incident wave frequency is 4.0 GHz. The direction of incident wave is . The polarization is . The size of the airplane is 52.4256 m × 21.0312 m × 2.944213 m. Electric size is about 700λ × 280λ × 39λ. The size of grid is m. The amount of total grids is 7020 × 2840 × 420 (about 8.4 billion). It is tested on Shanghai Supercomputer Center (SSC), using 10240 cores.
5.2.2. Result
The computation time of this code is 6036.86 seconds. The RCS of the airplane is shown in Figure 12.
(a) plane
(b) plane
(c) plane
6. Conclusion
A guideline is presented for using parallel FDTD on supercomputers with more than 10 thousands of CPU cores, based on a theoretical communication model given in this paper. The benchmarks obtained on two supercomputers validated the optimal virtual topology rules. Radiation from a large microstrip antenna array and scattering from an electrically large airplane are simulated successfully, which indicate the capability of the method presented in this paper for those types of reallife EM problems.
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
Acknowledgments
This work is supported by the National High Technology Research and Development Program of China (863 Program) (2012AA01A308), the NSFC (61301069, 61072019), the Project with Contract no. 2013KJXX67, and the Program for New Century Excellent Talents in University of China (NCET130949). The computational resources utilized in this research are provided by the National Supercomputer Center in Tianjin (NSCCTJ), National Supercomputing Center in Shenzhen (NSCCSZ), and Shanghai Supercomputer Center (SSC).
References
 A. Taflove, Computational Electrodynamics: The FiniteDifference TimeDomain Method, Artech House, Norwood, Mass, USA, 2000.
 D. Ge and Y. Yan, FiniteDifference TimeDomain Method for Electromagnetic Waves, Version 3, Xidian University Press, Xi'an, China, 2011, (Chinese).
 J. L. Volakis, D. B. Davidson, C. Guiffaut, and K. Mahdjoubi, “A parallel FDTD algorithm using the MPI library,” IEEE Antennas and Propagation Magazine, vol. 43, no. 2, pp. 94–103, 2001. View at: Publisher Site  Google Scholar
 U. Andersson, Timedomain methods for the Maxwell equations [Ph.D. thesis], Royal Institute of Technology, Stockholm, Sweden, 2001.
 Y. Zhang, J. Song, and C. H. Liang, “MPIbased parallelized locally conformal fdtd for modeling slot antennas and new periodic structures in microstrip,” Journal of Electromagnetic Waves and Applications, vol. 18, no. 10, pp. 1321–1335, 2004. View at: Publisher Site  Google Scholar
 Y. Zhang, J. Song, and C. Liang, “Study on the parallel modified locally conformal FDTD algorithm on cluster of PCs for PBG structures,” Acta Electronica Sinica, vol. 31, no. 12A, pp. 2142–2144, 2003. View at: Google Scholar
 Z. Yu, D. Wei, and C. Liang, “Analysis of parallel performance of MPI based parallel FDTD on PC clusters,” in Proceedings of the AsiaPacific Conference Proceedings: Microwave Conference Proceedings (APMC '05), vol. 4, December 2005. View at: Publisher Site  Google Scholar
 W. Yu, X. Yang, Y. Liu et al., “A new direction in computational electromagnetics: solving large problems using the parallel FDTD on the BlueGene/L supercomputer providing terafloplevel performance,” IEEE Antennas and Propagation Magazine, vol. 50, no. 2, pp. 26–44, 2008. View at: Publisher Site  Google Scholar
 http://www.nscctj.gov.cn/.
 http://www.nsccsz.gov.cn/.
 http://www.ssc.net.cn/.
 W. Chen and J. Zhai, Preliminary Analysis on Communication Performance of Dawning 5000A, 863 High Performance Computer Testing Center of Tsinghua University, 2008.
 Intel Corporation, Intel MPI Library for Linux OS Reference Manual, Intel Corporation, 2011, https://software.intel.com/sites/products/documentation/hpc/mpi/linux/reference_manual.pdf.
Copyright
Copyright © 2015 Shugang Jiang et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.