The Definitive Guide to SQL Server Performance Optimization

Chapter 1SQL Server uses caches to avoid not only physical reads but also time-consuming queryoptimization. Whenever SQL Server creates a query execution plan, it caches that plan for futureuse. Any subsequent uses of the exact same query won’t require a run through the queryoptimizer; instead, SQL Server will use the cached execution plan. SQL Server’s caches caninclude:• Stored procedures• Prepared statements• Ad-hoc queries• Replication procedures• Triggers• Views• Defaults• User tables• System tables• Checks• RulesEach of these objects can be stored in SQL Server’s caches for improved performance.Understanding Hardware Performance ComponentsIn the end, all performance comes down to hardware. Although database administrators oftenfind themselves tweaking queries for better execution times, they do so only because it’s easierthan upgrading their server hardware. In fact, most of the performance tuning that you do as adatabase developer or administrator will be to make your applications run faster on the hardwarethat you have. That doesn’t mean your applications or queries are poorly written; it simply meansthat you have to cut performance corners to make them run better on your hardware.Because all performance problems eventually come down to hardware, you need to take the timeto understand how SQL Server interacts with the hardware on which it runs. Doing so will helpyou select better hardware for new SQL Server computers and help you better utilize thehardware on existing SQL Server computers. From a performance point of view, there are fourbasic hardware components: disks, networks, processors, and memory.Disk I/OAs I’ve already mentioned, disk throughput is often the bottleneck on SQL Server computers.Even if you’re running 8GB of RAM on a server, your databases can easily consume onehundred times that much space, meaning SQL Server is doomed to constant disk access. Thereare, however, some tricks to improving disk I/O. I’ll show you how to identify and solve disk I/O problems in Chapter 3. Plus, in Chapter 6, I’ll explainhow you can scale out SQL Server to help solve disk I/O problems.Imagine, for example, that you need to produce a handwritten copy of a 1000-page book. Youcan either use one helper so that the two of you will have to each copy 500 pages, or you can usenine helpers so that you’ll each be responsible for only 100 pages. Because you can each copy afixed number of pages in any given period of time, the more hands you have helping you, the9