Multiprocessing, Research Paper Example

Introduction

Multiprocessing involves installation two or more processors in a single computer system. In doing this, there are different multiprocessor architectures that can be considered. The most common architectures are the master-slave multiprocessing (MSMP), symmetrical multiprocessing (SMP) and separate supervisor. Taking into consideration the MSMP and SMP, there are certain comparisons that can be made between the two from different dimensions.

Comparing and Contrasting the MSMP and SMP

The Master-Slave Multiprocessor (MSMP) Architecture

In this architecture, there is a processor that acts as a master to all the other processors. This processor is enabled to execute operations of the operating system, which the other processors cannot do. All the tasks of the slave processors are scheduled and controlled by the master processor (ACSAC et al, 2007). This implies that data structures are stored in its private memory. As the master processor executes operations of the operating system, the slave processors are left to take control of the application programs.

In this architecture, a single task can be performed in a parallel format by allocating a single task to two or more processors simultaneously. However, most of the input/output resources are only accessible by the master processor with limited permission given to selected slave processors. This system is relatively easy to design since it is only the master processor that executes the entire operating system. More emphasis is laid on the allocation of tasks to slave processors.

It mainly suffers the challenge of limited scalability since the master processor will in most cases fail to utilize all the resources of the slave processors. This arises from the fact that the multiprocessor may become a bottleneck in the course of allocation and execution of tasks.

Symmetrical Multiprocessing (SMP) Architecture

As opposed to MSMP, in SMP, all processors are equally configured. This brings about autonomy and equal treatment of all the processors in terms of allocation of tasks and resources of the computer system, including input/output resources (Chevance, 2005). Resources are pooled and made equally accessible to all the processors. The operating system is also made symmetrical so that all the processors are able to utilize it fully. This means that there is a single copy of the kernel that can be equally executed by all the processors simultaneously.

The design of this multiprocessor system thus requires to be controlled for proper interlocks necessary to access scarce information and data as well as the pooled resources. At this point, a similarity comes with the MSMP where only a single processor is allowed to access the multiprocessor at any given time. This is referred to as ‘floating master’ technique as a result of only one operating system being in operation amidst many processors. At any given time when a single processor executes the operating system, it acts as a master processor to all the remaining processors.

The main difference from the MSMP is that in the case of the MSMP, there is a particular processor that executes the operating system. In the MSP, the operating system is not bound to a specific processor, thus keeps floating from one processor to another, depending on the tasks that are being performed by all the processors. Parallel execution is enabled by aligning ready processors within a shared memory. Allocation is done step by step from the first processor to the last until all the processors are occupied for the tasks that are available.

Hyper-threading

Execution of multiple threads of tasks at the same time is the essence of multiprocessing. The architectures that are selected in most cases should be in a position to execute multiple threads at the same time. In the symmetrical multiprocessing structure, allocation of tasks is done in a particular pattern until all the processors are occupied. This implies that all the processors will be executing different sets of tasks at the same time, what encompasses multithreading.

In the MSMP, since the master processor is the only one that has access to the operating system, all the other processors are allocated to application programs. In this architecture, hyper-threading capabilities are minimized since a single task may be assigned to two or more processors. This implies that more than one processor will be allocated to perform a single task. This is a limiting factor of the MSMP. However, depending in the number of slave processors that are available for utilization, multi-threading capabilities may be incorporated into this architecture (Gerber & Binstock, 2004).

Applications for which MSMP is Suited

One of the areas where this architecture is highly applicable is in computer networking. In this application, there is only one device or process, referred to as master that controls all the other devices. One processor will be used to act as the server and regulates all the activities of the other processors. The slaves will be used to perform other activities within the computer network (Chevance, 2005). In a networking model, the MSMP structure has been put into application for a long time by the designers. Limitation of capabilities of the other systems gives the master processor all the capabilities to control the operations taking place within the network.

Industrial communications have greatly applied the master-slave model in such a way that the host computer has unlimited control over the communication channel. In this case, all the other devices are assigned to perform other tasks in the communication system.

Application areas of the SMP

There are networking models that also put into use the SMP architecture. This is in a case where multiple CPUs are assigned to share a common board. Despite sharing of these resources, each of them acts independently. Such networks are do not put limitations to any devices as far as sharing of resources is concerned, thus the SMP becomes highly applicable.

Scaling applications highly put into use the SMP multiprocessing. As a result of its multi-threading capabilities, the SMP architecture is also applicable in other areas such as storage caching algorithms, layer 3 packed forwarding and Internet Protocol route calculations (Sebastien 2010).

Conclusion

Increasing computer speed is necessary for faster performance of tasks. The concept of multiprocessing is an important area to be considered for various applications that require high performance speeds. Selecting a particular model to be used in multiprocessing depends on the type of tasks to be performed by the computer system in place. In the case of a networking system, a master-slave architecture would be the most appropriate to be considered. However, there are tasks that can take in any of the two discussed architectures. Taking into consideration these two models, the importance of multiprocessing can be evaluated and put into realistic applications. In addition to the two models, there are other multiprocessor architectures that can be put into use.

References

ACSAC 2007, Choi, L., Paek, Y., & Cho, S. (2007). Advances in computer systems architecture: 12th Asia-Pacific Conference, ACSAC 2007, Seoul, Korea, August 23-25, 2007: proceedings. Berlin: Springer.

Chevance, R. J. (2005). Server architectures: Multiprocessors, clusters, parallel systems, Web servers, and storage solutions. Amsterdam: Elsevier/Digital Press.

Gerber, R., & Binstock, A. (2004). Programming with Hyper-threading technology. Hillsboro, Or: Intel Press.

Sebastien, M. M. (2010). Using Symmetric Multiprocessing (SMP) to scale Data Plane and Control Plane Performance. QNX Software Systems