Parallel Kirchhoff Pre-Stack Depth Migration on Large High Performance Clusters
详细信息 查看全文
- 关键词：High performance computing ; Parallel computing ; Kirchhoff Pre ; Stack depth migration ; Seismic imaging ; Parallel algorithms
- 刊名：Lecture Notes in Computer Science
- 出版年：2015
- 出版时间：2015
- 年：2015
- 卷：9530
- 期：1
- 页码：251-266
- 全文大小：2,042 KB
- 参考文献：1.Perrone, M. P., Zhou, H., Fossum, G., Todd, R.: Practical VTI RTM. In: 72nd EAGE Conference and Exhibition, June 2010
2.Perrone, M., Liu, L. K., Lu, L., Magerlein, K., Kim, C., Fedulova, I., Semenikhin, A.: Reducing data movement costs: scalable seismic imaging on blue gene. In: 2012 IEEE 26th International Parallel & Distributed Processing Symposium (IPDPS), pp. 320–329. IEEE May 2012
3.Schmuck, F.B., Haskin, R.L.: GPFS: A shared-disk file system for large computing clusters. In: Proceedings of FAST 2002, vol. 2, p. 19 January 2002
4.Yilmaz, Ö., Doherty, S.M.: Seismic Data Processing, vol. 2. Society of Exploration Geophysicists, Tulsa (1987)
5.Chang, H., VanDyke, J.P., Solano, M., McMechan, G.A., Epili, D.: 3D pre-stack Kirchhoff depth migration: from prototype to production in a massively parallel processor environment. Geophys. 63(2), 546–556 (1998)CrossRef
6.Panetta, J., de Souza Filho, P.R., da Cunha Filho, C.A., da Motta, F.M.R., Pinheiro, S.S., Pedrosa, I., de Albrecht, C.H.: Computational characteristics of production seismic migration and its performance on novel processor architectures. In: 19th International Symposium on Computer Architecture and High Performance Computing, SBAC-PAD 2007. pp. 11–18. IEEE October 2007
7.Li, J., Hei, D., Yan, L.: Partitioning algorithm of 3D pre-stack parallel Kirchhoff depth migration for imaging spaces. In: Eighth International Conference on Grid and Cooperative Computing, GCC 2009. pp. 276–280. IEEE August 2009
8.Bevc, D.: Imaging complex structures with semi-recursive Kirchhoff migration. Geophys. 62(2), 577–588 (1997)CrossRef
9.Yin, Y., Byna, S., Song, H., Sun, X.H., Thakur, R.: Boosting application-specific parallel i/o optimization using IOSIG. In: 2012 12th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGrid), pp. 196–203. IEEE May 2012
10.Yu, H., Sahoo, R.K., Howson, C., Almasi, G., Castanos, J.G., Gupta, M., Gropp, W.D.: High performance file i/o for the blue Gene/L supercomputer. In: The Twelfth International Symposium on High-Performance Computer Architecture, pp. 187–196. IEEE February 2006
11.Dai, H.: Parallel processing of pre-stack Kirchhoff time migration on a PC cluster. Comput. Geosci. 31(7), 891–899 (2005)CrossRef
12.Sen, V., Sen, M.K., Stoffa, P.L.: PVM based 3-D Kirchhoff depth migration using dynamically computed travel-times: an application in seismic data processing. Parallel Comput. 25(3), 231–248 (1999)MATH CrossRef
13.Marr, D.T., Binns, F., Hill, D.L., et al.: Hyper-threading technology architecture and microarchitecture. Intel. Technol. J. 6(1), 4–15 (2002)
14.Nowak, D.A., Seagar, M.: ASCI tera-scale simulation: requirements and deployments. http://www.ornl.gov/sci/optical/docs/Tutorial19991108Nowak.pdf
15.Heien, E., Kondo, D., Gainaru, A., LaPine, D., Kramer, B., Cappello, F.: Modeling and tolerating heterogeneous failures in large parallel systems. In: 2011 International Conference for High Performance Computing, Networking, Storage and Analysis (SC), pp. 1–11. IEEE November 2011 - 作者单位：Chao Li (17)
Yida Wang (17)
Changhai Zhao (18)
Haihua Yan (17)
Jianlei Zhang (18)
17. School of Computer Science and Engineering, Beihang University, Beijing, 100191, China
18. Research and Development Center, BGP Inc., CNPC, Zhuozhou, 072751, Hebei, China - 丛书名：Algorithms and Architectures for Parallel Processing
- ISBN：978-3-319-27137-8
- 刊物类别：Computer Science
- 刊物主题：Artificial Intelligence and Robotics
Computer Communication Networks
Software Engineering
Data Encryption
Database Management
Computation by Abstract Devices
Algorithm Analysis and Problem Complexity - 出版者：Springer Berlin / Heidelberg
- ISSN：1611-3349
文摘
Kirchhoff Pre-Stack Depth Migration (KPSDM) is a widely used algorithm for seismic imaging in petroleum industry. To provide higher FLOPS, modern high performance clusters are equipped with more computing nodes and more cores for each node. The evolution style of clusters leads to two problems for upper layer applications such as KPSDM: (1) the increasing disparity of the I/O capacity and computing performance is becoming a bottleneck for higher scalability; (2) the decreasing Mean Time Between Failures (MTBF) limits the availability of the applications. In this paper, we present an optimized parallel implementation of KPSDM to adapt to modern clusters. First, we convert the KPSDM into a clear and simple task-based parallel application by decomposing the computation along two dimensions: the imaging space and seismic data. Then, those tasks are mapped to computing nodes that are organized using a two-level master/worker architecture to reduce the I/O workloads. And each task is further parallelized using multi-cores to fully utilize the computing resources. Finally, fault tolerance and checkpoint are implemented to meet the availability requirement in production environments. Experimental results with practical seismic data show that our parallel implementation of KPSDM can scale smoothly from 51 nodes (816 cores) to 211 nodes (3376 cores) with low I/O workloads on the I/O sub-system and multiple process failures can be tolerated efficiently.