Computer cluster

Introduction

cluster system typically referred to as nodes, are typically connected via a LAN, but there are also other possible connections. Cluster computers typically used to improve the calculation speed and / or reliability of a single computer. Group computer than a single computer, such as a workstation or a supercomputer much higher cost performance under normal circumstances.

cluster classification

and homogeneous cluster into two kinds of isomers, their difference is: Architecture between computer components are the same cluster. Group computer according to the functions and structures can be divided into the following categories:

High Availability Cluster High-availability (HA) clusters

Load balancing cluster Load balancing clusters

HPCC High-performance (HPC) clusters

grid computing Grid computing

high availability cluster

generally refers to the case where the cluster when a node fails, the tasks which are automatically transferred to other normal node. Can also refer to a node in the cluster offline for maintenance on the line again, the process does not affect the operation of the entire cluster.

cluster load balancing

cluster load balancing typically run through one or more front end load balancer to distribute the workload on a set of back-end server, to achieve high overall system performance and high availability. Such a computer cluster also sometimes called a server farm (Server Farm). General high availability clustering and load balancing cluster will use similar technology, or both high availability and load balancing features.

Linux Virtual Server (LVS) project provides the most common load balancing software on the Linux operating system.

HPCC

HPC clusters using computing tasks assigned to a different cluster of computing nodes children increased computing power, and thus is mainly used in the field of scientific computing. The more popular HPC Linux operating system and other free software to complete parallel computing. This cluster configuration is often called Beowulf clusters. Such clusters typically runs a specific program to play HPC cluster parallel capabilities. Such procedures are generally application-specific runtime, such as MPI library specifically designed for the computing science. Computationally intensive job data communications

HPC are particularly suitable for clusters between each computing node in the calculation, intermediate results such as a node or nodes other circumstances affecting the calculation result.

grid

or a mesh grid computing cluster is one kind of cluster very relevant art. The main difference with the traditional grid computer cluster is a group of related grid connection is not trusted, it operates more like a computing utility, rather than a stand-alone computer. Also, the grid is usually more than a cluster of different types of computer support collections.

Grid computing is for many tasks for independent work of optimization, in the calculation process without sharing data between jobs. Grid mainly in the service management job assignment between computers independent execution of the job. The storage resource can be shared by all nodes, but the intermediate results of the job does not affect the progress of the job in the other grid nodes.

Cluster Technology

1, with the completion of the work by a plurality of computers. To achieve higher efficiency.

2, the contents of the two machines or machine, and the like exactly the same working process. If a crash, the other can work.