随着地震数据规模的不断增大,在进行数据并行处理时,并行计算通信框架因使用TCP(Transmission Control Protocol)协议存在网络吞吐量低、高时延等性能问题,以及主从并行模式下存在主节点网卡性能瓶颈问题,制约了数据并行处理效率的线性...随着地震数据规模的不断增大,在进行数据并行处理时,并行计算通信框架因使用TCP(Transmission Control Protocol)协议存在网络吞吐量低、高时延等性能问题,以及主从并行模式下存在主节点网卡性能瓶颈问题,制约了数据并行处理效率的线性提升,集群节点规模扩展性下降明显。为此,提出采用RoCE(RDMA over Converged Ethernet)协议替换TCP协议、主节点配置高性能100 GE网卡的高效并行计算通信优化方案,解决了并行计算框架存在的数据网络传输性能问题及主节点同时收集多个计算节点计算结果数据的网络性能瓶颈问题,实现了高速可扩展技术的大规模地震数据处理通信应用方案,计算节点可快速完成数据通信,提升了大规模地震数据处理计算效率;另外采用UCX(Unified Communication X)技术还提升了应用系统的可移植性与使用的便捷性。逆时偏移处理数据的测试结果表明,对于本次大规模数据的处理,逆时偏移并行计算效率提升了32.8%,效果显著,可缩短大规模地震数据逆时偏移计算的时间,并减少计算能源消耗,因此具有很高的实用价值和经济效益。展开更多
This paper reports a method to make hazard maps of sediment disasters resulting from an earthquake and following heavy rainfall for the entire region of Gunma prefecture, Japan. Firstly, we identified the slopes in th...This paper reports a method to make hazard maps of sediment disasters resulting from an earthquake and following heavy rainfall for the entire region of Gunma prefecture, Japan. Firstly, we identified the slopes in the study area, which are susceptible to large-scale landslides and land failures during an earthquake with a magnitude of seven on the Richter scale. To analyze the sheer volume of the data, we employed a statistical method to evaluate the susceptibility, mainly considering geomorphologic conditions. Secondly, we extracted mudflow and slope failure susceptible areas and potential flooding zones resulting from a damming at a river triggered by the earthquake and heavy rainfall, and we identified the settlements which would be isolated by the road disruption caused by the sediment disasters. As the result, 359 settlements were classified as potential isolation areas. Combining the above-mentioned susceptibility maps, we obtained two types of sediment disaster hazard maps of the study area, depicting the potential hazards which would occur during the earthquake and the disasters which would be caused by heavy rainfall following the quake, respectively. These hazard maps and the disaster information would be useful for the regional disaster prevention planning and countermeasures in the future.展开更多
The paper begins with a brief review of the research history of earthquake size measurement. On this basis,the author pointed out the following points: 1 In recent decades,ML,mb( mB),MSmagnitude scales are widely used...The paper begins with a brief review of the research history of earthquake size measurement. On this basis,the author pointed out the following points: 1 In recent decades,ML,mb( mB),MSmagnitude scales are widely used as measures of earthquake size. However,these magnitude scales have a deficiency of "overgeneralization " and"magnitude saturation". Moreover,since they do not fully take into account the regional difference of seismic attenuation, especially the difference of site effects on the amplification of ground motion,these magnitude scales are but inaccurate measures of earthquake size. 2 Seismic moment M0 not only has clear physical meaning,but also overcomes the deficiencies existing in ML,mb( mB) and MSmagnitude scales,so it is the most suitable physical quantity for measuring earthquake size scientifically. In order to continue to use the term "magnitude",Kanamori defined the moment magnitude scaleΔσ. Although its prerequisite assumptions remain to be studied,it is still a reasonable scale used as a measure of the relative size of an earthquake. 3 For measuring the earthquake size more scientifically,we must make full use of a large amount of waveform data from modern regional digital seismograph networks,strengthen the research on seismic wave attenuation characteristics,site effect,calculation of source parameters and the related scaling relations. In improving the measurement methods for ML,mb( mB) and MSmagnitude,we should focus on the improvement of MWscale and carry forward the work as gradually taking MWmagnitude scale as the uniform physical quantity to measure the relative size of earthquakes,so as to lay a more solid foundation for research in earthquake science and earthquake prediction.展开更多
文摘随着地震数据规模的不断增大,在进行数据并行处理时,并行计算通信框架因使用TCP(Transmission Control Protocol)协议存在网络吞吐量低、高时延等性能问题,以及主从并行模式下存在主节点网卡性能瓶颈问题,制约了数据并行处理效率的线性提升,集群节点规模扩展性下降明显。为此,提出采用RoCE(RDMA over Converged Ethernet)协议替换TCP协议、主节点配置高性能100 GE网卡的高效并行计算通信优化方案,解决了并行计算框架存在的数据网络传输性能问题及主节点同时收集多个计算节点计算结果数据的网络性能瓶颈问题,实现了高速可扩展技术的大规模地震数据处理通信应用方案,计算节点可快速完成数据通信,提升了大规模地震数据处理计算效率;另外采用UCX(Unified Communication X)技术还提升了应用系统的可移植性与使用的便捷性。逆时偏移处理数据的测试结果表明,对于本次大规模数据的处理,逆时偏移并行计算效率提升了32.8%,效果显著,可缩短大规模地震数据逆时偏移计算的时间,并减少计算能源消耗,因此具有很高的实用价值和经济效益。
文摘This paper reports a method to make hazard maps of sediment disasters resulting from an earthquake and following heavy rainfall for the entire region of Gunma prefecture, Japan. Firstly, we identified the slopes in the study area, which are susceptible to large-scale landslides and land failures during an earthquake with a magnitude of seven on the Richter scale. To analyze the sheer volume of the data, we employed a statistical method to evaluate the susceptibility, mainly considering geomorphologic conditions. Secondly, we extracted mudflow and slope failure susceptible areas and potential flooding zones resulting from a damming at a river triggered by the earthquake and heavy rainfall, and we identified the settlements which would be isolated by the road disruption caused by the sediment disasters. As the result, 359 settlements were classified as potential isolation areas. Combining the above-mentioned susceptibility maps, we obtained two types of sediment disaster hazard maps of the study area, depicting the potential hazards which would occur during the earthquake and the disasters which would be caused by heavy rainfall following the quake, respectively. These hazard maps and the disaster information would be useful for the regional disaster prevention planning and countermeasures in the future.
基金funded by the Basic R&D Special Fund of Institute of Earthquake Science,CEA(2012IES0204)
文摘The paper begins with a brief review of the research history of earthquake size measurement. On this basis,the author pointed out the following points: 1 In recent decades,ML,mb( mB),MSmagnitude scales are widely used as measures of earthquake size. However,these magnitude scales have a deficiency of "overgeneralization " and"magnitude saturation". Moreover,since they do not fully take into account the regional difference of seismic attenuation, especially the difference of site effects on the amplification of ground motion,these magnitude scales are but inaccurate measures of earthquake size. 2 Seismic moment M0 not only has clear physical meaning,but also overcomes the deficiencies existing in ML,mb( mB) and MSmagnitude scales,so it is the most suitable physical quantity for measuring earthquake size scientifically. In order to continue to use the term "magnitude",Kanamori defined the moment magnitude scaleΔσ. Although its prerequisite assumptions remain to be studied,it is still a reasonable scale used as a measure of the relative size of an earthquake. 3 For measuring the earthquake size more scientifically,we must make full use of a large amount of waveform data from modern regional digital seismograph networks,strengthen the research on seismic wave attenuation characteristics,site effect,calculation of source parameters and the related scaling relations. In improving the measurement methods for ML,mb( mB) and MSmagnitude,we should focus on the improvement of MWscale and carry forward the work as gradually taking MWmagnitude scale as the uniform physical quantity to measure the relative size of earthquakes,so as to lay a more solid foundation for research in earthquake science and earthquake prediction.
