Hybrid domain parallel algorithm for 3D Kirchhoff prestack depth migration in the heterogeneous environment
Wang Yida1, Zhao Changhai2, Li Chao1, Zhang Jianlei2, Yan Haihua1, Zhang Weiyi2
1. School of Computer Science and Engineering, Beihang University, Beijing 100191, China; 2. Research & Development Center, BGP Inc. CNPC, Zhuozhou, Hebei 072751, China
Abstract:The 3D Kirchhoff prestack depth migration (KPSDM) is the most important depth-domain imaging method in the seismic data processing.Currently seismic data size of a single survey exceeds 100TB,and will increase to more than 1PB in the near future.Considering the continuous increasing survey sizes and the introduction of programmable graphic process unit (GPU),the conventional parallel strategy is no longer appropriate for the large-scale heterogeneous processing clusters.In this paper,we propose a practical hybrid domain parallel KPSDM algorithm based on two-level decomposition including imaging space and seismic data.The algorithm eliminates the dependency among tasks.In a heterogeneous environment,we implement the computing part on GPU and design a "dynamic and asynchronous" task allocation policy to achieve load balancing on heterogeneous computing system.Because KPSDM,as part of its execution,usually requires repeated access to huge seismic data and a large amount of travel time tables,the scalability is always limited by the shared storage maximum throughput.To solve the scalability problem,we build a distributed cache system using the local storage for a KPSDM job spans.It can provide a very high aggregate data bandwidth to supply seismic data and travel time table to a running task timely.The KPSDM implementation can obtain close to linear speedup when it processes real seismic data on a 256-node cluster.
Addair T G,Dodge D A,Walter W R et al.Large-scale seismic signal analysis with Hadoop.Computers & Geosciences,2014,66(5):145-154.
[2]
罗刚,陈继红,孙孝萍等.大规模异构集群地震作业调度与资源管理系统的设计与实现.石油地球物理勘探,2017,52(增刊2):200-205.Luo Gang,Chen Jihong,Sun Xiaoping et al.Heterogeneous cluster scheduling and resource management system for a large number of seismic data processing jobs.OGP,2017,52(S2):200-205.
[3]
荣骏召,芦俊,李建峰等.矢量Kirchhoff叠前深度偏移.石油地球物理勘探,2017,52(6):1170-1176.Rong Junzhao,Lu Jun,Li Jianfeng et al.Vector pre-stack depth migration based on Kirchhoff integral equation.OGP,2017,52(6):1170-1176.
[4]
Gao Y,Iqbal S,Zhang P et al.Performance and power analysis of high-density multi-GPGPU architectures:A preliminary case study.IEEE 17th International Conference on High Performance Computing and Communications (HPCC),2015,66-71.
[5]
Jack D,Erich S,Horst S et al.Top 500 List.http://www.top500.org/,2017.
[6]
Shi X,Li C,Wang S et al.Computing prestack Kirchhoff time migration on general purpose GPU.Computers & Geosciences,2011,37(10):1702-1710.
[7]
NVIDIA.Cuda C Programming Guide.http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html,2013.
[8]
赵长海,罗国安,张旭东等.大规模异构集群上Kirchhoff叠前时间偏移并行算法.石油地球物理勘探,2016,51(5):1040-1048.Zhao Changhai,Luo Guoan,Zhang Xudong et al.Kirchhoff prestack time migration on large heterogeneous computing systems.OGP,2016,51(5):1040-1048.
[9]
Chang H,Van Dyke J P,Solano M et al.3-D prestack Kirchhoff depth migration:From prototype to production in a massively parallel processor environment.Geophysics,1998,63(2):546-556.
[10]
王华忠,刘少勇,孔祥宁等.大规模三维地震数据Kirchhoff叠前深度偏移及其并行实现.石油地球物理勘探,2012,47(3):404-410.Wang Huazhong,Liu Shaoyong,Kong Xiangning et al.3D Kirchhoff PSDM for large-scale seismic data and its parallel implementation strategy.OGP,2012,47(3):404-410.
[11]
Rastogi R,Srivastava A,Khonde K et al.An efficient parallel algorithm:Poststack and prestack Kirchhoff 3D depth migration using flexi-depth iterations.Computers & Geosciences,2015,80(7):1-8.
[12]
Panetta J,de Souza Filho P R P,Da Cungua Filho C A et al.Computational characteristics of production seismic migration and its performance on novel processor architectures.19th Symposium on Computer Architecture and High Performance Computing (SBAC-PAD 2007), IEEE,2007,11-18.
[13]
Panetta J,Teixeira T,de Souza Filho P R P et al.Accelerating time and depth seismic migration by CPU and GPU cooperation.International Journal of Parallel Programming,2012,40(3):290-312.
[14]
Li J,Hei D,Yan L.Partitioning algorithm of 3-D prestack parallel Kirchhoff depth migration for imaging spaces.Eighth International Conference on Grid and Cooperative Computing,2009,276-280.
[15]
Cunha C A,Pametta J,Romanelli A et al.Compres-sion of traveltime tables for prestack depth migration.SEG Technical Program Expanded Abstracts,1995,14:180-183.
[16]
Alkhalifah T.The many benefits of traveltime compression for 3D prestack Kirchhoff migration.68th EAGE Conference & Exhibition Extended Abstracts,2006.
[17]
Alkhalifah T.Efficient traveltime compression for 3D prestack Kirchhoff migration.Geophysical Prospecting,2011,59(1):1-9.
[18]
Teixeira D,Yeh A,Gajawada S.Implementation of Kirchhoff prestack depth migration on GPU.SEG Technical Program Expanded Abstracts,2013,32:3683-3686.
Li C,Wang Y,Yan H et al.High performance Kirchhoff pre-stack depth migration on Hadoop.Procee-dings of the Symposium on High Performance Computing,Society for Computer Simulation International,2015,158-165.
[21]
Schroeder B and Gibson G A.A large-scale study of failures in high-performance computing systems.IEEE Transactions on Dependable and Secure Computing,2010,7(4):337-350.
[22]
Cappello F,Geist A,Gropp W et al.Toward exascale resilience:2014 update.Supercomputing Frontiers and innovations,2014,1(1):5-28.
[23]
赵长海,晏海华,王宏琳等.面向地震数据的并行与分布式编程框架.石油地球物理勘探,2010,45(1):146-155.Zhao Changhai,Yan Haihua,Wang Honglin et al.Seismic data processing oriented parallel and distributed programming framework.OGP,2010,45(1):146-155.
[24]
Dean J and Ghemawat S.MapReduce:simplified data processing on large clusters.Communication of the ACM,2008,51(1):107-113.
[25]
Qi C,Cheng L and Zhen X.Improving MapReduce performance using smart speculative execution strategy.IEEE Transactions on Computers,2014,63(4):954-967.