2.11 Summary

< Free Open Study >

A computer system is comprised of many elements. Primarily these are the central processing unit, the memory unit, the secondary storage unit, the input and output unit, and interconnection hardware. Each of these elements can be architected in a variety of ways, each with their own set of pros and cons. Computer systems are typically represented as either single processor units, multiple processing units, distributed processing units, or networked units. The policies for connecting these devices to meet the needs of an application will dictate the final form of the system's architecture.

The operating system and related support infrastructure services are used by an application to organize, maintain, and manipulate information on a specific computer architecture. The applications and operating system's needs and priorities do not always match. Due to this impedance mismatch, applications and have, in the past, tried to work around the operating system instead of working with it. The most notorious was in the early days of the IBM PC and the DOS operating system. Application programs typically bypassed the operating system and worked directly on the underlying hardware. The result was that programs typically ran "IBM PC or 100% compatible" machines. Another example is the operating system's management of the memory hierarchy may be fair and reasonably optimal for the average application running on the system but may not match the needs of the database management system. The operating system strives to maintain a reasonable set of data pages in memory for the application's use, but it does not attempt to go beyond its own measures of effectiveness. The concept today is to engineer systems so that they operate optimally, based on the semantic needs and intent of the applications, which may go against the operating system's average response time and fairness goals.

< Free Open Study >