Application of the hottest high performance comput

  • Detail

The application of high-performance computer in petroleum industry

Abstract: high performance computing technology has always been a competitive hotspot in the world, and cluster technology is an important implementation method of today's high-performance computing technology. This paper focuses on the application of cluster technology in the petroleum industry, takes the trucluster cluster of Compaq company as an example, analyzes it, and prospects the advantages and application prospects of cluster technology

key words: cluster trucluster CFS network MPI

I. The importance of high-performance computing: as we all know, the development of modern science depends on three important means, theoretical research, scientific experiments and computing technology. The integration of the three leads to the rapid development of modern science. The development of computing technology has led to the deepening of theoretical research, the acceleration of scientific experiments and the reduction of costs. Today, with the increasingly fierce international high-tech competition, high-performance computing, which is at the highest end of the computing field, will still be the commanding height in the IT field in this century. High performance computing technology is highly valued at home and abroad. The development of science and technology, economy, society and enterprises has an endless demand for high performance computing technology. The high-performance computer that provides high-performance computing has the strongest function, the highest performance and the largest scale in all kinds of computer systems. It has huge numerical computing and data processing capabilities, and can be widely used in the national economy, national defense construction and scientific and technological development. Such as oil exploration, earthquake prediction, satellite image processing, genetic engineering and so on. The application of high-performance computer in petroleum industry is mainly manifested in scientific calculation, image processing, artificial intelligence, exploration and development database establishment, etc. As far as seismic exploration is concerned, the main method of oil and gas resource exploration in modern petroleum industry is three-dimensional seismic exploration. The collected kinematic and dynamic information of seismic waves propagating in the stratum is processed to remove the false and retain the true, and transformed into geological information. After calculation and analysis, the underground rock stratum and lithologic structure are deduced. These programs not only have large amount of calculation, long processing cycle, but also occupy a lot of storage space. Therefore, ordinary computers are not competent. Therefore, seismic data processing, reservoir numerical simulation, well logging data processing and research in the oil industry all need the support of high-performance computers

second, the advantages of clustering:

according to the architecture, there are four main types of high-performance computer systems at present. The first type is multi vector processing system, such as Cray YMP 90; The second is symmetric multiprocessor (SMP) system based on shared memory, such as SGI power challenge; The third type is the large module parallel processing (MPP) system based on distributed storage. Nanjing xuankai company has not only built the domestic leading polyglutamic acid and polylysine production lines, such as Intel Paragon; The fourth type is the cluster computer system formed by high-speed interconnection. Because the cluster system has the characteristics of low investment risk, good scalability, inheritable existing software and hardware resources, short development cycle, good programmability and so on. Compared with the traditional parallel processing system, it has so many obvious advantages. The parallel programming platforms of the above four types of high-performance computers are mainly divided into three categories: message passing (MPI), shared storage and data parallelism. Among them, message passing has good portability, and programming is simple and convenient. For example, the parallelization of the program is only to insert the corresponding communication primitive (MPI message passing library function) into the original C, c++ or FORTRAN serial program. Users still use the familiar programming environment and do not need to adapt to the new environment, so they can inherit the original software resources. So it is the most popular in the world at present, and cluster is also the most suitable for programming with message passing. In the process of developing the above-mentioned high-performance computing system, people also gradually realize that the scale scalability and programmability of the system have become the key issues for the further development of high-performance computing system. According to experts' prediction, "the future high-performance computers and super servers will be based on the cluster system structure". At present, the cluster system has become a high-performance, but due to the addition of conditioning beam, it can be calculated as a hot spot and mainstream. At present, representative cluster systems include IBM's SP2, Dec's trucluster, sun's SPARC cluster 1000/2000pdb, etc. in a broad sense, a cluster is independent of each other. Some systems work as a system, and these systems are connected by a high-speed general network according to a certain structure. They form a group called cluster. The purpose of forming a cluster is to improve the availability and scale scalability of the system. In the cluster system, each server undertakes its own tasks, so the number of servers will increase in direct proportion to the work efficiency. For programmers and users, the cluster system is a whole parallel system. The host and network in the cluster system can be isomorphic or heterogeneous. Compared with the current representative cluster systems, Compaq trucluster cluster has the best performance. Advantages of cluster system: most cluster systems have the following two advantages: availability: when an unexpected failure occurs to a server or application in the cluster, the cluster software will respond quickly and assign the tasks of the system to other working servers in the cluster, that is, to provide non-stop services

scalability: that is, scale scalability. When the 2017 China Belarus new materials industry seminar and the China Belarus industrial park new materials project docking meeting are jointly hosted by the Belarusian Academy of Sciences, the China Belarus Industrial Park Management Committee and the China Belarus Industrial Park Development Co., Ltd. in Beijing, and the overall work flow exceeds the capacity of each system in a cluster, other systems will be added to the cluster, thereby enhancing the overall performance of the system. In 2013, the global plastic output exceeded 280 million tons. In such a configuration, multiple servers can perform the same application and data operations

III. application of trucluster in oil enterprises:

the most prominent features of trucluster include:

1. CAA (cluster application availability) facilities provide the ability of resource monitoring and application restart for the cluster system. Its great advantage is that it can not only monitor the system resources as a member of the cluster, but also monitor the actual cards, so that the system administrator can define the failover or reallocation policy of services, and set the application running script. The CAA framework enables administrators to easily integrate an application into the trucluster environment, making an application highly available in just a few minutes

2. single system image technology

in order to make the cluster system different from the distributed system, the cluster system must have a single system image (SSI) to some extent, which means that when facing a cluster, users feel a unified system. In terms of single system image technology, the core technologies of Compaq trucluster include

Copyright © 2011 JIN SHI