Denormalization and advanced indexing techniques are highly effective for improving the performance of your data warehouse and other applications. But, regardless of how effectively you use these techniques, you will still need to frequently execute large and complex queries. To get good query response times, you need (among other things) high performance and scalability out of your I/O subsystem. By using scalable and parallel I/O techniques, you can push your system components (disks, controllers, CPUs and memory) to the limits of their capacity, thereby fully utilizing your hardware's capabilities.

One of the key factors in a database I/O discussion is the disk reader process, commonly referred to as a "scan thread" or "scan process," which is responsible for getting data off of a disk and sending it to a CPU. It is important to realize that at any given point in time, a scan thread can only read from a single disk and can only ship that data to a single CPU. In addition, each disk can only serve a single request from a single scan thread at a time. Therefore, if we have four CPUs in our system, unless we have four scan threads and at least four disk drives, all four CPUs will not be able to run simultaneously--at least one would be idle.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access