What Are Shared-Memory Multiprocessors?

While computer hardware was often expensive and fairly unobtainable for the standard consumer during the technology’s infancy, prices have since seen a steady drop leading into the present. Now, consumers have much easier access and ability to create more complex and powerful systems with common components available on the market. With a number of consumer motherboards now offering more than one slot for CPU attachment, shared-memory processors can be used to achieve higher system performance for a number of applications.

Shared-memory processors are a type of system that contains multiple processors that may carry out their operations together. Through a shared interconnection network, the processors can utilize the same pool of memory and communicate with one another to carry out various procedures. As such, computers with shared-memory processors can exhibit a significant difference in their computation power as compared to standard work stations with only one processor. As these assemblies are typically geared more towards demanding applications and processes that may require large amounts of program execution, many casual users may not find much use in running a shared-memory processor set-up.

In the case of an internet, database, or network server, however, having the most processors possible is paramount to smooth operations and ensuring that the servers are able to accommodate periods of high usage and user loads. Additionally, shared-memory processors can also serve to streamline certain applications, as a computer system can utilize large amounts of power to conduct a single job rather than computing a high number of small jobs at the same time. When connecting processors together, each processor is joined from their independent data caches to a shared memory pool through a single interconnection network.

Known as symmetric multiprocessing hardware, such components allow for the assembly and pairing of multiple processors so that each CPU has equal control over memory and peripherals. Across most symmetric multiprocessing hardware assemblies, buses and crossbars serve as the primary method for interconnection. In regard to computer hardware, a bus is a component that allows for data to be transferred, and they are commonly seen on many motherboards for the connection of memory, CPUs, and more. A crossbar, on the other hand, is a component containing a series of switches that may be used to conduct information processing applications. Out of the two symmetric multiprocessing hardware pieces, the bus serves as the most convenient and common approach for establishing a shared-memory multiprocessor assembly. With the bus, connections for parts, protocols, and hardware are all provided to facilitate operations with ease. As buses are limited in their ability to handle high amounts of data traffic, it is important that loads do not exceed the performance standards of the bus as to avoid bottlenecking.

With the use of a crossbar, bottlenecking is avoided as multiple paths may operate simultaneously on a grid-like system. As an example, a 4x5 crossbar could allow for up to four active data transfers to be conducted at the same time. By having a higher number of active paths as compared to a singular shared bus, more performance can be achieved. While these advantages are clearly desirable, crossbar components typically range much higher in price, and their cost only increases as the load raises. Due to this, crossbars are mostly reserved for the most high-end applications.

1. steven helmer
2. Posted on January 20, 2021
aviation