We live in a data-centric world that consumes information at an amazingly fast pace. The information we process as individuals all starts out as data stored somewhere and, more importantly, is a requirement driven by an application that generates the information that we are presented with. Where did all these applications come from and why do they require such vast amounts of data? Certainly these are questions that are beyond the scope of this book; however, they remain the fundamental driving force behind the innovation and advancement of I/O processing.
The support of the increasing data-centric nature of OLTP and data analysis application systems evolved through experiences in large database applications using centralized mainframe configurations. Even as some high-end processor configurations became distributed by connecting systems within a tightly coupled network, the ability to handle high I/O transactional rates and large-scale databases exceeded their limitations. The advancement and proliferation of relational database solutions exacerbated the problem with its exponential storage resource appetite when compared to traditional processing of the time. As this relational database phenomenon filtered into the client/server processing configurations, this increased not only the number of server configurations, but their complexities and challenges as well. It also provided another target to support the ever increasing population of data-centric applications.
Many of today's applications rely on some form of relational database products that, as discussed in Chapter 3, were designed to work directly with disk storage devices. In addition to this evolution of processing, combining large user populations with Very Large Data Bases (VLDB) requires a greater sophistication of I/O and storage functionality than what is available on traditional, albeit large-scale, mainframe, client/server storage systems, or NAS.
The effects of the high end OLTP and data-centric data warehouse applications accelerated the advancement of I/O architectures within high-end server systems and mainframes. These also spawned the development of increasingly complex distributed processing configurations to support the growth of I/O operations driven by business requirements that called for increasing amounts of data to be available online. These advancements have taken the form of Symmetric Multiprocessing Systems (SMP) and Massive Parallel Processing (MPP) systems. These architectures advanced the I/O capabilities and functionality of traditional computer systems, and as a lasting effect set a foundation for the development of storage area networking.