Cluster
A cluster is a number of networked computers, i.e. a computer network. To handle extensive tasks, individual computers are combined to form a network, which then forms a unit. Hardware and software engineers are faced with enormous challenges in terms of compatibility when forming clusters. So-called “homogeneous clusters”, whose computers run under the same system and the same hardware, have a simpler structure. “Heterogeneous clusters,” on the other hand, may use different operating systems and hardware. The set of computers (also called “nodes” or “servers”) in a cluster is often referred to as a “server farm.” The use of clusters comes into play when dealing with mainly three different tasks:
-
- an increase in availability combined with fail-safety
- load distribution over several computers
- an increase in performance to cope with computationally intensive jobs
Load-balancing cluster
A load-balancing cluster is a computer network that is optimized for load distribution. When large amounts of data need to be processed, they are distributed among individual computer stations. This protects the system from total failures. For many companies, load-balancing clusters are an inexpensive solution because they save them from having to purchase expensive special computers. The system automatically recognizes which computers in the network are less busy and assigns new incoming tasks to them accordingly. The increase in capacity is not achieved by upgrading individual computers, but by integrating additional computers.
High-performance computing clusters
HPC clusters or high-performance clusters are used for particularly computationally intensive tasks. They are often used in the scientific field. The tasks are either executed in parallel by several nodes or they are distributed among the individual nodes. The distribution of the tasks is usually handled by a so-called “job management system”. The first cluster product for load sharing came onto the market as early as 1977, but was not able to establish itself. In 1983, the product VAXCluster appeared, which enabled parallel computing on the nodes as well as the shared use of file systems. It is still marketed today by HP as “VSMCluster”.
