Grid environment for computational astrophysics driven by GRAPE-6 with HMCS-G and OmniRPC

Grid environment for computational astrophysics driven by GRAPE-6 with HMCS-G and OmniRPC
复制标题

DOI:
10.1109/ipdps.2005.236
复制
发表时间:
2005-04
期刊:
19th IEEE International Parallel and Distributed Processing Symposium
影响因子:
--
通讯作者:
T. Boku;K. Onuma;M. Sato;Yoshihiro Nakajima;D. Takahashi
T. Boku;K. Onuma;M. Sato;Yoshihiro Nakajima;D. Takahashi
中科院分区:
其他
文献类型:
--
作者:
T. Boku;K. Onuma;M. Sato;Yoshihiro Nakajima;D. Takahashi

文献摘要

被引文献

相似文献

在计算天体物理学中,引力的影响是必不可少的,即使是非常繁重的计算O(N/sup 2/)的N粒子。一些特殊用途的机器已经被实现为重力发动机,以非常快地处理这个问题,但是世界上很少有站点操作这样的系统。我们已经开发了一个网格环境访问这样一个系统的基础上网格RPC,命名为HMCS-G。使用HMCS-G,可以与PC集群和重力引擎GRAPE-6的组合进行多物理计算天体物理模拟。然而,在典型规模的星系形成问题中,PC集群的计算能力远低于GRAPE-6,我们需要多组PC集群来共享网格环境下GRAPE-6的计算能力。我们已经开发了这样一个系统使用OmniRPC,网格启用RPC系统。在该系统中,OmniRPC被用来分配作业的参数搜索在多个PC集群和访问GRAPE-6服务器从这些集群。我们在该系统上进行了一个实际问题的求解,并证实了这种解对于各种计算天体物理研究都是有用的。
In computational astrophysics, the effect of gravity is essential even with very heavy computation of O(N/sup 2/) for N particles. Several special purpose machines have been implemented as gravity engine to handle this problem extremely fast, however there are few sites in the world to operate such systems. We have developed a grid environment to access such a system based on grid-RPC, named HMCS-G. Using HMCS-G, a multi-physical computational astrophysics simulation can be performed with the combination of a PC-cluster and the gravity engine GRAPE-6. In typical size of problems for galaxy formation, however, the computation power of PC-cluster is much weaker than that of GRAPE-6, and we need multiple sets of PC-clusters to share the power of GRAPE-6 on grid environment. We have developed such a system using OmniRPC, a grid-enabled RPC system. In this system, OmniRPC are used both to distribute jobs for parameter search on multiple PC-clusters and to access GRAPE-6 server from these clusters. We performed an actual problem to search several formations of galaxy on this system, and confirmed such a solution is useful for wide variety of computational astrophysics research.