October 5th, 2021

Machine translation is used partially for this article. See the Japanese version for the original article.

Introduction

Now, let's continue learning from the previous article.

Components in the server can avoid becoming SPOF (Single Point of Failure) with various redundancy techniques, but the OS itself can not.
In other words, redundancy seems to have a limit on a single server.
So this time, let's learn about redundancy with multiple servers.

Redundancy with multiple servers is called clustering.
Load-balancing clusters and HA clusters are commonly used to protect systems from failure.

Load balancing cluster

Load balancing cluster is commonly used on Web servers and are configured as follows:

When accessing a web server, the client accesses the load balancer instead of accessing the web server directly.
Access from multiple clients is distributed and processed to each server by the load balancer.
It is possible to distribute the load by dividing and processing accesses from clients among multiple servers.

This allows a single server to handle as much accesses than a single server can provide.
Even if one server fails, as long as the other servers are operating normally, the system will be able to handle the accesses from clients.

HA cluster

HA stands for "High Availability".
"Availability" means the ability of a system to operate continuously, and HA cluster can enhance this ability.

HA clusters are commonly used on DB servers and the configuration is as follows.

In an HA cluster, applications (DB applications in this case) never run on all servers at the same time.
If an application is running on one server, the application is down on the other servers.
This is the most obvious difference from load balancing clusters.
Load balancing clusters accommodate access allocated from load balancing devices on each server. Therefore, it is assumed that applications are running on all servers.

On the other hand, the HA cluster clearly distinguishes between a server running an application (active server) and a server that is not (standby server).
If a failure occurs on the server running the application, the other servers can take over the application to enhance availability.

So, what exactly does it mean to have another server take over the running of the application? Let's learn about it next time, so that's all for today.

Tips

You may be wondering, "In the end, which should I use, a load balancing cluster or HA cluster?"
In case you want to enhance the availability of the system, i.e., you don't want to stop the system even if the server fails, you need to choose a cluster configuration depending on the specifications of the application you are using and the purpose of the system.
First, consider the configuration of the load balancing cluster.
If the application you are using does not support load balancing clusters, then you should consider HA cluster configuration. However, since this depends heavily on the application specifications.
Therefore, you should ultimately check with the application vendor.