The Terminology Maze

In this article I have introduced the terms active redundancy and passive redundancy and related them to Assured Availability and High Availability respectively. In reading this article, you could begin to equate the word active with Assured Availability, and the word passive with High Availability. You might then be surprised if in further research you came across High Availability systems that call themselves “active/active” systems.

The words active and passive are used in another way when discussing the role of servers in dual or multi-server systems. The term active server refers to the server that is actively performing system processing. Passive server refers to the server that is not performing system processing, but is standing by waiting to take over upon failure of the active server. This is called an active/passive configuration.

Since this is inefficient and can be considered a wasteful use of the passive server, another approach is sometimes taken. Two servers capable of handling all system functions are set up to each perform half of the system processing (either by type of task or by volume of transactions). These servers then back each other up, and either one will take over for the other if the other fails. This is called an active/active configuration.

Active/active server configurations can exist in both Assured Availability systems and High Availability systems. The system descriptions can be misleading.

One product I read about was described as an “active/active continuously available system”, which sounds like Assured Availability to me. Later on in fine print it was explained that the system is “continuously available except for the brief time when the failover transition occurs”. [Emphasis added.] The system does not carry over the user logons or application states, and no mention at all was made about the failback transition time. While it is true that system is continuously available for those users who are logged on to the server that did not fail, it is not true for users logged into the failed server. How important this distinction is depends upon the type of system you are running.

Words used one way in this article can be found elsewhere used in a different way or for a different purpose. The main point is that you must be sure to find out exactly what is meant by the terminology describing fault tolerant products you investigate, in order to accurately evaluate and compare products. Different manufacturers may use the same words and terms in different ways.