Читать книгу Cloud Computing Solutions - Группа авторов - Страница 22

A computer cluster can be characterized as an arrangement of several coupled computers cooperating in such a way that every machine can be viewed as a single system image (SSI). Computer clusters are developed by merging a colossal number of computer developments, including access to fast networks, low-cost MPs, and software that delivers high-performance computing.

According to Sadashiv and Kumar [4], a cluster can be defined as the collection of distributed or parallel computers attached among themselves with the help of high-speed networks such as SCI, Myrinet, Gigabit Ethernet and InfiniBand. They function collectively in the execution of data- and compute-intensive tasks that would not be feasible for a single computer to execute alone. The clusters are mostly used for load balancing (to distribute the task over the different interconnected computers), high availability of the required data, and for compute purpose. The interconnected computers are used due to their high availability as they maintain the redundant nodes which are being utilized to convey the required service when the system components fail.

The performance of the system is upgraded enough and enhanced in that case because regardless of whether one node neglects to figure out the task, there is another backup node which will be ready to convey the task and takes on the simple single purpose without any snags [5]. At the point when numerous computers are connected in a computer cluster, they can easily share computational workload as a single virtual computer. From the client’s perspective, they are numerous machines, yet they are working as a single virtual machine. The client’s demand is received and appropriated among all the independent computers to shape a computer cluster. This outcome is adjusted and reasonable computational workload is shared among various machines, enhancing and improving the computational performance of the cluster systems. Frequently clusters are used for the most part for computational purposes, and than for taking care of IO-based exercises.