Common cluster types

Common cluster types

1. High Availability Cluster (High Availability Cluster) HA

Running on two or more nodes, the purpose is to minimize the service interruption time in the case of certain failures in the system, and to ensure the ability of the application to continue to provide services. The more famous such clusters are keepalived, TurbolinuxTurboHA, Heartbeat, Kimberlite, etc. There are many popular names for such clusters, such as "dual machine hot standby", "dual machine mutual standby" and so on.
A high-availability cluster means that the availability of services is relatively high. When one of our servers crashes, our services will not be unavailable. Its working mode is to transfer a faulty service to a normal working server, so that the service will not be interrupted. Generally speaking, the servers working in the front-end (distributor) of our cluster will perform a health check on our back-end servers. If we find that our servers are down, they will not be forwarded. The measurement standard: availability = online time/ (Online time + fault processing time), which is usually the availability of 99%, 99.9%, 99.99% and so on. Common open source solutions in the industry include heartbeat, keepalived, etc.;

2. Load Balance Cluster LB

Provides a load capacity proportional to the number of nodes. This type of cluster is suitable for services that need to provide large-load access, such as Web. The more famous such clusters are Turbolinux Cluster Server and Linux Virtual Server. This type of cluster reasonably distributes the load pressure to each computer in the cluster according to a certain algorithm to reduce the pressure on the main server and reduce the hardware and software requirements of the main server.
There is a distributor or scheduler in the load balancing cluster. We call it Director. It is located on top of multiple servers. The distributor selects one from the following server clusters according to internally defined rules or scheduling methods. In response to the request sent by the client, thereby solving the problem of high concurrency. In the expansion, it is very easy to use scale out to expand, so as to realize the scalability of the cluster. Common open source solutions in the industry include lvs, haproxy, nginx, ats;

3. High Performance Computing Cluster HPC

Use computing cluster software to connect the computers of multiple nodes together to complete computing tasks that usually only supercomputers can complete.
HP: A high-performance cluster is when a certain task is very large, we build a cluster to complete this task together. This processing method is called parallel processing cluster. Parallel processing cluster is a mechanism that divides large tasks into small tasks and processes them separately. It is often used for big data analysis and massive resource integration. Currently, Hadoop is more famous.
In short, in the actual production environment, we can choose the appropriate cluster solution according to the actual situation. The three types of clusters have different focuses. The LB cluster focuses on providing concurrent service processing capabilities, and the HA cluster improves the online service capabilities to achieve services. Uninterrupted, the HP cluster is focused on processing a massive task