Introduction to Cluster Computing Presentation:

Cluster Applications: The different types of applications on the cluster have different behavior. Some require high bandwidth and low latency capabilities. Some need commodity cluster. Between these two ends, many messaging algorithms and applications fall. There are four types of applications namely compute intensive, data or input/output intensive, and transaction intensive. Each have unique characteristics. Computing intensive applications are those which require a lot of computations.

These applications are very sensitive latency in messages. These applications are also graphic intensive. In data intensive applications large storage is required. Performance of these applications is dependent on the architecture of the cluster, the bandwidth of the network attached storage. Data intensive applications are beneficial in data mining, genome, and protein science applications. In transaction intensive applications, there is high interaction between application resources and cluster resources. This is used in banking, financial, human resources, and web based applications.

Performance Impacts:

Here, first Data throughput calculation of theoretical maximum throughput is done and then effective throughput is calculated which is always less than the theoretical throughput. Throughput for clusters is based on total no. of nodes, switch architecture, forwarding technology, queuing methodology, buffering depth and allocation, noise and error on the cable plant.

Slow start:

In the original implementation of TCP, when the connection was established between two devices, they could send segments as fast as they liked till there is space in other device’s receive window. In busy network, new traffic can create congestion. To solve this problem the modern TCP devices are restricted at the rate at which they send their segments. In starting they can send only a limited amount of segments. Then each time they receive an acknowledgement they can send more segments. Thus the device starts slow.

Download  Cluster Computing Presentation .