Sorry to add some confusion to the discussion, but...
It is absolutely unclear for me what is it that author was trying to achieve in his comparison?
Reading the second sentence I have found that "This article outlines the differences between them, for the purposes of maintaining a standby server"
It might be my misunderstanding, but I am not familiar with such a business requirement? I was always thinking about a high availability which is in order a part of a wider DR policy which could be a business requirement and standby server as a very narrow methodology which is a part of suggested high availability solution?
The very next sentence is adding to my personal confusion:
"To put this in context,log-shipping and replication are two methods which are often compared to clustering, however clustering is really a technology created to solve a different business requirement - that of automatic fail-over."
Automatic fail-over is not a business requirement as well. It is a part of functionality that could be listed as a partial requirement in the same HA policy. By itself automatic fail over means nothing... More than that I can implement clustering without fail over at all.
MS suggesting following HA solutions that includes (bolded for SQL 2005):
* SQL Server Database – Failover Clustering, Mirroring, Log Shipping, Replication
* SQL Server Analysis Services – Failover Clustering
* SQL Server Notification Services – Failover Clustering
Replication is the lowest choice in the tree of possible HA solutions. If functional replication requirements were included in the original design of the database, it is possible to recommend it as a partial HA solution.
Excluding clustering from the list of possible solutions for SQL 2000 is a mistake. Based on the author's words, one can think that clustering is implemented for automatic fail over only. At the same time a comparison of a top level functionality of clustering and replication is showing that cluster implementation will require less resources and what is most significant less administrative overhead (especially on complex SQL server installation), AND provide a MORE full HA solution than replication.
Personally I am not a fan of clustering solution. However, when making my recommendations to the customer, I am trying to keep in mind that this is just a tool. 🙂
In my understanding a picture of comparison of HA tools will not be finished without bringing mirroring and ESPECIALLY third party tools (like XOSoft, for example).