Category parallel computing, distributed computing, grid computing and cloud computing belong HPC (High Performance Computing, HPC), the main purpose is to analyze and process large data, but they present many differences. ENC cloud services are based on theoretical performance computing techniques, by improving the traditional services architecture for fast access and processing operations massive electronic chart data, it is better in the area of marine Sciences GIS compute-intensive problems and provide data-intensive computing and processing capabilities. High-performance computing architecture is the foundation technology of electronic chart massive data storage and processing, but also smooth prototype electronic chart cloud service assurance system development. The following will parallel computing, distributed computing, difference and connection between the grid and cloud computing for analysis.
Parallel computing
Parallel computing is calculated relative to the serial, it refers to a calculation model that allows multiple instructions simultaneously, parallel in time and space can be divided into parallel. I.e., parallel in time using the number of lines while working, the space refers to the use of parallel processors to perform a plurality of concurrent computations to reduce the time required to solve complex problems. From the application developer's perspective, parallel computing can be divided into parallel data and functional parallelism, data parallelism to achieve the same operation in parallel subtasks by decomposition of the data, the parallel function to achieve the same data by different tasks of the task decomposition of parallel jobs. In comparison, the parallel data easier to achieve, so this will also be provided based on the principle of parallel data in the parallel algorithm. For the study of parallel computing started in the 1970s, it has the relevant theory, research parallelism multiple data streams (the MIMD) as single instruction multiple data stream (SIMD) and multiple instruction, 1980s in structure Parallel Architecture with great results, there have been multiple computers using the network of parallel structures and the use of shared memory multi-processor parallel computer composed of scientists used this parallel computing architecture to reduce the time to solve complex problems in the field of sophisticated technology.
Can be drawn from the above analysis, early parallel computing is mainly used in scientific research, with a specific application environment, need to use high technical skills to complete the task parallel programming required. Although at the time parallel computing from P
Through public is still far away, but has to solve complex problems (such as functional parallelism, data parallelism, communication and coordination) laid the methodological foundation. It can be said, it is the initial stage of parallel computing cloud or is infancy, it provides a practical and simple ideas and basic ideas for the development of cloud computing.
Distributed computing
Distributed computing is a need to perform compute-intensive engineering data partitioned into small pieces and processed by multiple networked computers, respectively, after uploading the results, the results of unified scientific data merge draw conclusions. In the 1990s, with the final to determine the TCP / IP protocol, the rapid development of computer network, Web Service and other network new technology comes as a wide area network-based distributed computing hardware and software to do the foundation. First, the similarities and differences compared Distributed Computing and Parallel Computing. The similarities are that simplify complex tasks into multiple sub-tasks, then operating simultaneously on multiple computers. The difference is that distributed computing is a more loosely structured, less demanding real-time, across LAN deployments running on the Internet, a large number of public projects (such as a black hole exploration, drug discovery, protein structure analysis, etc.) they use this embodiment, the parallel computing is the need for communication between nodes, a strong correlation between the node more frequently, mainly deployed in the LAN through the high-speed network. In distributed computing algorithm, we are more concerned about communication rather than the algorithm steps between computers, distributed computing because communication cost compared to a single node on the right to influence the overall performance of much greater importance.
Can be derived from the above analysis, it is the product of a distributed computing network development, is calculated by the parallel evolve new modes: parallel computing network. If the parallel computing to cloud computing laid the theoretical foundation, it was distributed computing cloud computing has laid a solid technical cornerstone of the network.
Grid
Grid computing refers to the use of computer resources by a plurality of independent entities or institutions large heterogeneous (processor cycles and disk storage), a standardized uniform and open interfaces and access protocols, non-centralized control access of resources of the formula and collaborative problem solving, in order to achieve the quality of service is higher than the sum of the system each member of the grid system cumulative quality of service.
After the mid-1990s, to a certain stage of development of distributed computing, grid computing began, its purpose is to use decentralized network resources to solve computationally intensive problems. At that time, high-end computer hardware expensive, researchers have tried to define special protocol mechanisms to enable network resource management to decentralized heterogeneous and dynamic, to solve the problem of high-end computer-intensive operations to solve. Grid virtual organization concept and thereby generating, by defining a series of standard protocols, and a middleware toolkit to implement allocation and scheduling of resources in virtual organizations. Its focus is to support cross-domain capability and heterogeneous computing resource integration, making it simple and traditional computer cluster or distributed computing phase difference. To enable grid computing to become routine public services like water grid, Ian Foster proposed acquisition should define a protocol or standard computing storage resources in the network, under the guidance of this theory, the world organization has designed a series of network grid system, such as OSG, ESG, EGEE, these grid computing system capable of providing even more data storage resources owned services and functions in accordance with the requirements of the service required of the designer. OASIS, OGF, etc. ISO has also developed standards, grid computing was once considered the cluster computing market. So far, however, the commercial grid system still does not appear. The concept is too large, very complex protocol standards so that Grid project truly practical behavior are driven by the state, such as EUGrid, DataGrid, ChinaGrid, EduGrid and so on. However, the development of grid computing, provides the basic framework for network support for emergence of cloud computing.
Cloud computing
Cloud computing is a large data storage analysis and by the elastic expanding and contracting demand-driven resource calculation model, by which a virtualization, dynamic scale pool of resources to provide users with high-availability, efficiency, calculation of the elastic memory resources
Service data function. Comprising five key features: ① distributed parallel computing techniques; ② to achieve scale, elasticized computation storage; ③ virtual users and services with multiple stages; ④ by high-performance computing and large data storage drive; ⑤ dynamic service resources, elasticized. The reason cloud computing in recent years to obtain widespread concern mainly the following three points: ① decrease costs and enhance the computing power storage device, multi-core, multi-processor technology and the popularity of the birth; ② the industry has accumulated more and more professional data, need effective use; ③ widely used network services and applications Web2.0.
From the above analysis, cloud computing in parallel with a conceptual level, computing clusters, grid computing, distributed computing intersection exists, a cloud as described is not only to calculate the evolution of the grid, and the grid computing also
provides the basic framework for network support cloud computing. The focus of grid computing is to provide computing and storage capabilities, and more focused on abstract cloud resources and service capabilities, which is grid computing to cloud computing evolution. Compared with distributed computing, cloud computing is a mature and stable flow of business resources, it provides an abstract service amount calculated for the user just as hydropower plants provide water resources amount calculated as convenient and reliable. Figure 1.1 shows the relationship with other relevant concepts of cloud computing. Web2.0 interpretation of service-oriented development, cloud computing has become the main force of them; parallel computing and cluster computing program to focus more on traditional application-oriented design; the concept of grid computing because of its huge and these four areas are crossed, from a broad perspective, distributed computing domain comprising the whole concept.
Based on the above analysis, we can conclude that the relationship between these concepts. From the perspective of the user computer, parallel computing is done by a single user, distributed computing is done by a multi-user collaboration, grid computing is performed by calculating a large heterogeneous tissue, the cloud without user involvement is the other end of an elastic service cluster complete.
Content from Information Engineering University, Dr. Liu Canyou papers.
----------------
Disclaimer: This article is the original article CSDN bloggers "GeoWin_CAS", and follow CC 4.0 BY-SA copyright agreement, reproduced, please attach the original source link and this statement. .
Original link: https: //blog.csdn.net/yaoxiaochuang/article/details/41542293
