Server Redundancy Architecture

When designing server infrastructure, server redundancy is configured for stable service operation during runtime.

By configuring physical or logical servers, or LAPR, etc., if one service fails, the service can continue through another server.

Server redundancy configurations can typically be divided into Active-Active and Active-Stand by configurations.

Active-Active #

An Active-Active architecture is configured to distribute processing between server 1 and server 2 based on function and nature, using load balancing from an L4 switch or similar (SLB Server Load Balancer).

In addition to web servers, two servers are also configured when needed for database servers.

Most web servers are configured with L4 switch SLB, and DB servers utilize Oracle RAC (Real Application Cluster).

Disk sharing can also be configured similarly with Veritas CFS (Cluster File System) or similar.

This configuration allows operations to continue separately on server 1 and server 2 in case of a specific device failure, and there is no downtime.

Active-Stand By #

Servers are configured for redundancy, but load balancing is not simultaneously served to all devices.

This means it's configured to transfer and operate services in case of a failure. (failover)

When the operating system (main server), commonly referred to as the production system, experiences a failure, it recognizes the service outage and transfers the service to a secondary server.

This process periodically checks the system's normal status using cluster heartbeats, etc.

What is a cluster heartbeat? It's a periodic signal sent between nodes in a cluster to check if each node is functioning correctly (health check). It plays a crucial role in maintaining stability and availability within the cluster.

If an unusual situation occurs, services are manually switched to the secondary server through a system engineer's decision, or automatically switched in the event of a critical failure.

Ultimately, the purpose of such server redundancy is to prevent unavoidable failures and minimize downtime.