Overview

Symmetrical multiprocessors made their appearance during the 1960s in the high- end mainframe space, driven by the need for scalability. As we have explained earlier, scalability is the property that makes it possible to adapt a system to the size of the problem to be handled, and while problem size is often equated to the size of the processing problem it should not be forgotten that other dimensions are just as important, storage requirements and communications being key examples.

The processing capability of a symmetrical multiprocessor is adjusted to meet needs simply by installing additional processors until the requirements are met. This allows a customer to increase the capabilities of an installed system by simple addition of processor modules, rather than having to replace the whole machine with a higher-performance version. In a symmetrical multiprocessor, the workload is distributed fairly among all the installed processors.

For simplicity, we will refer to symmetrical multiprocessors by the industry-standard acronym, SMP. SMPs are also referred to as tightly-coupled multiprocessors.

The SMP approach is an economically-attractive way of implementing a range of systems; the power of the various models in the range is characterized not only by the capabilities of the processors used (for example, internal microarchitecture and maximum frequency) but also by the maximum number of processors that can be installed.

After its introduction in the mainframe world, the approach was adopted by minicomputer manufacturers (in the 1970s) and then in the 1980s for UNIX systems based on standard microprocessors.