ACKNOWLEDGMENT This work is partly supported by the Science Funds of China U0635004 and No. 60571056, and the Science Funds of Guang- dong 07118061. REFERENCES 1. Y.-S. Lin, C.-C. Liu, K.-M. Li, and C.-H. Chen, Design of an LTCC tri-band transceiver module for GPRS mobile applications, IEEE Trans Microwave Theory Tech 52 (2004), 2718 –2724. 2. J.-T. Kuo, T.-H. Yeh, and C.-C. Yeh, Design of microstrip bandpass filters with a dual-passband response, IEEE Trans Microwave Theory Tech 53 (2005), 1331–1337. 3. S.-F.R. Chang, W.-L. Chen, S.-C. Chang, C.-K. Tu, C.-L. Wei, C.-H. Chien, C.-H. Tsai, J. Chen, and A. Chen, A dual-band RF transceiver for multi-standard WLAN applications, IEEE Trans Microwave Theory Tech 53 (2005), 1048 –1055. 4. M.H. Weng, H.W. Wu, and Y.K. Su, Compact and low loss dual-band bandpass filter using pseudo-interdigital stepped impedance resonators for WLANs, IEEE Microwave Wireless Compon Lett 17 (2007), 187– 189. 5. Y.P. Zhang and M. Sun, Dual-band microstrip bandpass filter using stepped-impedance resonators with new coupling schemes, IEEE Trans Microwave Theory Tech 54 (2006), 3779 –3785. 6. C.H. Lee, C.I.G. Hsu, and H.K. Jhuang, Design of a new tri-band microstrip BPF using combined quarter-wavelength SIRs, IEEE Micro- wave Wireless Compon Lett 16 (2006), 594 –596. 7. H. Zhao and T.J. Cui, Novel triple-mode resonators using split-ring resonator, Microwave Opt Technol Lett 49 (2007), 2918 –2922. 8. X.M. Lin and Q.X. Chu, Design of triple-band bandpas filter using tri-section stepped-impedance resonators, in Proc Int Conf Microwave Millimeter Wave Tech 2007, pp. 798 – 800. © 2008 Wiley Periodicals, Inc. PERFORMANCE ANALYSIS OF PARALLEL FDTD METHOD ON THE DIFFERENT PLATFORMS Wenhua Yu, 1 Yongjun Liu, 1 Xiaoling Yang, 1 Raj Mittra, 1 Yongquan Lu, 2 Pai Wang, 2 Qing Che, 2 and Rui Lu 2 1 Electromagnetic Communication Laboratory, EE East 319, The Pennsylvania State University, PA 16801 2 High Performance Computing Center Communication, University of China, Beijing 100024; Corresponding author: wxy6@psu.edu Received 2 January 2008 ABSTRACT: In this communication, we investigate the performance of parallel FDTD code on a Gigabit Ethernet and optical network. It is fact that the optical network is necessary for a large cluster due to its wide bandwidth and small latency. However, for a small cluster, the Gigabit Ethernet will have a better ratio of performance to price. We also investigate the difference between high performance server and regular PC for the parallel FDTD code. © 2008 Wiley Periodicals, Inc. Microwave Opt Technol Lett 50: 2465–2467, 2008; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/mop. 23661 Key words: FDTD; gigabit ethernet; optical network; parallel code 1. INTRODUCTION Compared to method of moments (MoM) [1] and finite element method (FEM) [2], FDTD method [3, 4] is more suitable for the parallelism due to its unique update procedure. Parallel FDTD method allows us to use multiple processors with distributed memory to simulate a large EM problem. The field update inside each processor does not require any information from its neigh- bors, however, the field updates on the interface between proces- sors need the information at both the sides. The MPI is a library not a type of computer language, and it has become an international standard. The MPI functions are employed to pass the field infor- mation from one processor to its neighbors. Although the MPI library includes more than 200 functions, but the basic MPI functions are only six, namely, MPI_Init (spawn processes on the CPU’s) and MPI_Finalize (kill all remote processes), MPI_Comm_size (Return number of processes), MPI_Comm_rank (return the number of pro- cess and rank), MPI_Send (send a message), MPI_Recv (receive a message). MPI [5] is a library and has different implementation. The original implementation of MPICH is called MPICH1 and it im- plements the MPI-1.1 standard. As of 2006, the latest implemen- tation is called MPICH2 and it implements the MPI-2.0 standard. MPICH is a kind of middleware and realize the MPI functions on the different computer platforms so the developers can concentrate on their own code development. In this communication, we inves- tigate the performance of Gigabit Ethernet and optical network using the FDTD simulation for a small cluster. We also compare the performance of a high performance server and a regular PC with and without large I/O in the FDTD simulations. 2. NUMERICAL EXPERIMENTS Networks used in a computer cluster can be divided into two categories, namely, gigabit Ethernet and optical network. Both gigabit Ethernet and optical network are complicated systems; the detailed discussion is beyond the topic of this communication. What we care in the parallel processing is their bandwidth and latency. However, in the parallel FDTD method, we only need to pass the field information on the interface between subdomains, Figure 7 Photograph of the fabricated triple-band microstrip filter. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com] DOI 10.1002/mop MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 50, No. 9, September 2008 2465