Chapter 9: SQL Server 2005 High Availability

Overview

There are a number of dimensions that you need to keep in mind when considering investment and rollout of large or very busy systems: First, every system has its point of failure (POF). In the last chapter, we examined those points. It dealt with issues affecting storage, processors, memory-as well as the bottlenecks that can materialize at those points.

Second, performance and availability come at a price. Ergo, POF has meaning In monetary terms as well…the point at which the budget goes supernova. There is a point at which the cost consumes available funding and consumes all generated revenue from sales or investors.

Third, linear scalability, especially of processors, is a myth. There is a point at which throwing more processors at a system no longer has any advantage and may even begin to degrade performance. This is the point at which vertical scalability tapers off and horizontal scalability takes over, becoming not only desirable but also more reliable and less risky.

Fourth, there is no such thing as 100 percent availability of any single system. If there is one thing more certain to a good DBA or system administrator than death, taxes, and a headhunter around the next corner, it is that there will come a time when your server goes up in smoke. How you plan for that event is what’s important. Some people run away from fires; others run for a bag of marshmallows. This brings us to a fifth dimension-manageability.

The more complex and delicate a system is to manage, the higher the cost of management and the higher the risk of system failure or loss. We have discussed a number of graphical tools in recent chapters, so we will not tackle more here, save for the interfaces for installing onto clusters, creating federations and partitions, and so on.

In this chapter, the largest f or the most complex operational responsibility, well examine some availability theory, service level (SL), and availability and scalability solutions specific to SQL Server 2005. We will then wrap up the chapter with a step-by-step cluster exercise, as well as a recovering from a complete and utter disaster.

Note 

One gigabyte (GB)=1,024 megabytes (MB); one terabyte (TB)=1,024 GB; one petabyte (PB)=1,024 TB; one exabyte (EB)=1,024 PB.