A cluster is a group of computers that provide a group of network resources to users as a whole. These individual computer systems are the nodes of the cluster. An ideal cluster is that users never realize the underlying nodes of the cluster system. In their view, the cluster is a system rather than multiple computer systems. And administrators of the cluster system can freely add and delete nodes of the cluster system.
Clustering is not a completely new concept. In fact, as early as the 1970s, computer manufacturers and research institutes began to research and develop cluster systems. Because they are mainly used for scientific engineering calculations, these systems are not well known to everyone. Until the advent of Linux clusters, the concept of clustering was widely disseminated.
Research on clusters has its origins in the good performance scalability of cluster systems. Increasing the CPU clock speed and bus bandwidth is the primary means of providing computer performance. However, this method provides limited system performance. Then people increase performance by increasing the number of CPUs and memory capacity, so there are vector machines, symmetric multiprocessors (SMPs), etc. However, when the number of CPUs exceeds a certain threshold, the scalability of multiprocessor systems such as SMP becomes extremely poor. The main bottleneck is that the CPU's bandwidth for accessing memory cannot effectively increase as the number of CPUs increases. Contrary to SMP, the performance of a cluster system changes almost linearly with the number of CPUs.

Several computer system extensibility
 
The advantages of a cluster system are not only here, but here are the main advantages of the cluster system:
High scalability: as described above.
High availability: A node in the cluster fails and its tasks can be passed to other nodes, effectively preventing single point of failure.
High-performance: Load-balanced clusters allow the system to access more users at the same time.
Cost-effective: High-performance systems can be constructed using inexpensive industrial-standard hardware.
 
2.2 Introduction to High Performance Computer Clusters
 
In short, High-Performance Computing is a branch of computer science that focuses on developing supercomputers, studying parallel algorithms, and developing related software. High-performance computing is mainly applied to the following two types of research.
* Large-scale scientific issues like weather forecasting, topographic analysis and biopharmaceuticals
* Store and process massive data like data mining, image processing and gene sequencing
As the name implies, high-performance clusters use cluster technology to study high-performance computing.
 
There are many factors that affect the classification of high-performance computer systems, so the system architecture of high-performance cluster computers derived from different classification standards is also different. However, considering the use of commercially-available mass-produced computers and mass-produced commercial LANs and ATM networks, the following two architectures are mainly used.
At present, mainstream PC servers generally contain one to four processors, and access to a unique memory address space under the control of an operating system. Each processor is equal in memory access, and the access cost is the same. So this system is called a uniform memory access system, or SMP (Symmetric Multi-Processing Computer System) system. In this system, the processor is connected to the memory with a system bus, crossbar switch, or other proprietary network.
The biggest characteristic of the SMP system is: On the one hand, the system architecture and the shared memory programming model are relatively simple, so the manufacturing and application costs are very low, and are favored by the majority of SMEs. This is the main reason we use to build cluster nodes. On the other hand, its scalability is relatively poor. Generally, if the number of processors exceeds four, its manufacturing cost will become considerable. This restricts its further development.
However, in order to make up for this deficiency, the industry has introduced another system, a distributed memory access system. Its biggest characteristic compared with SMP system is that there is not a unified memory space. Each processor of this system and its memory space form an independent system, controlled by an operating system, and can operate independently. Here we call it a node, which uses its network interface to connect to the Internet and communicate through the network.