Internet and Distributed Application Services - Essay Example

?Explain Internet and Distributed Application Services Explain Internet and Distributed Application Services The term cluster is used in the computer technology to refer to a group of loosely coupled computers that are configured to work as a unit (Englander, 2003). Every computer within the cluster often operates as a complete unit and not as a tightly coupled multiprocessing system. Clustered computers or nodes have independent Central Processing Unit (CPU), I/O, and memory (Englander, 2003). Components of clustered computers are mostly interconnected using local area networks and Ethernet connection between other connection forms (Englander, 2003). Different forms of clusters exist and each has its failures, and advantages that paper intends to examine. Introduction The term cluster is used in the computer technology to refer to a group of loosely coupled computers that are configured to work as a unit. Every node often operates as a complete unit as opposed to the operations of a tightly coupled multiprocessing system. Clustered computers usually have independent CPU, I/O, and memory (Englander, 2003). Clustered computers may constitute multiple processing systems. A computer in a cluster refers to a node (Englander, 2003). The main aim of clustering computers is to make the users to see them as single machines, which they are not in the actual sense. There are the main reasons for clustering computers; however, each reason has its challenges and better share of benefits. The main advantages of clustering include increasing the computing power of computers by combining the individual power of each computer. Notably, a computer has the power to process data independently (Englander, 2003). Thus, increasing computers will automatically increase the nodes due to the increase in the number of computers thereby making the processing power of the cluster to be intense or higher than the processing power of single or individual computer or node within the cluster. Brewer among other contributors stated that clustering is inherent, absolutely, and incrementally scalable; therefore, installation adds nodes incrementally and this is pegged on the computer power needed for the intended cluster (Englander, 2003). The incrementally constituted clusters are often powerful and usually operate at low costs compared to even situations where single large machines are used. Fundamentally, clustering is a significant practice in computer technology since it is a sure way of increasing or designing highly performing computer systems. The processing problem including parallel processing units can be solved by breaking the problem into subtasks and then distributing them to different or parallel processing units among the nodes thereby solving the problem in parallel (Englander, 2003). Clustering also intends to create fault tolerant systems because each computer with the cluster is able to operate alone without the assistance from other computers within the cluster (Englander, 2003). Therefore, a node failure will not lead to standstill operational since the entire cluster will not be closed. The software controlling the entire cluster is capable of switching processes to other nodes within the cluster in a process referred to failover. Therefore, failures of certain nodes, the failed nodes’ processes can be switched to other functional nodes within the system; this switching will facilitate the continuation of processes that were running on the failed node or computer (Englander, 2003). Nonetheless, a cluster that has never experienced any form of failure can be designed and this is often advantageous for systems that process critical applications. Clusters are also vital in creating high available systems where computers within the cluster can be distributed geographically over a wide area (Englander, 2003). Therefore, users can access the closest computer system with the cluster. In most cases, this phenomenon creates a natural balancing of loads among different or series of computer nodes within the cluster. Clustering software has managed to balance processing workload equally across different nodes even beyond. In fact, the distribution of workload in cases of any failure within is an important feature that allows shifting of load to other nodes or computers within the cluster in case of system failure, for instance power failure. Furthermore, the system’s backup in increased (Englander, 2003). Clustering also allows load balancing especially where there are large workloads. Load balancing is the division of system’s workloads into constituent units to simply processing and storage; for instance, email addresses or accounts for large institutions or organizations can be divided alphabetically and be assigned to different machines for processing and storage. This clustering property allows effective and efficient data processing with minimal cases of system failure (Englander, 2003). Clustering has two fundamental models including shared nothing and shared disk modes. The shared nothing model is similar to the point-to-point network connection of computers where each computer is assigned its own disc. However, the difference is in the availability of a high-speed message to links between the clustered computers or the nodes (Englander, 2003). Moreover, there is system has a software the controls the interaction between nodes and behavior of each node. In this connection model, at least one computer will be designated to provide access to the outside world as well as managing the other nodes or the cluster (Englander, 2003). Therefore, the link passes data and messages between the nodes. Furthermore, the same link can partition the disks as per the changes in the conditions (Englander, 2003). The partitioning is done as demanded by data and messages by each node and it must be done equally (Englander, 2003). This makes the connection method viable and advantageous since little data sharing is experienced between the nodes. The main challenge of this model is that it has difficulties in predicting accurate partitioning (Englander, 2003). Thus, a particular node may be under or over utilized thereby affecting the efficiency of the entire cluster. Another model is the shared disk model that allows sharing of data between the cluster nodes since the disks are accessible equally to all nodes (Englander, 2003). The main advantages of this mode are that it is dynamic to workload balancing, has high availability, and is fault tolerant. Moreover, its availability is enhanced through many systems including the use of RAID technology between the shared disks (Englander, 2003). Regardless of these advantages, the shared disk cluster is costly due to the complexity in its software. The Beowulf clusters are highly configurable and simple clusters that are designed to provide highly performing computer systems at lower costs (Englander, 2003). These clusters have of multiple or many computers that are interconnected by private Ethernet that is dedicated to link servers between computers within a cluster (Englander, 2003). However, the clusters can be designed and configured in the form of shared nothing or shared disk model (Englander, 2003). Models in the Beowulf cluster have a memory and CPU as well as an Ethernet connection. They also contain floppy and hard disks among other peripherals (Englander, 2003). The Beowulf clusters are in most cases configured using either of the two computer components including COTS and Blade. COTS also known as commodity off the shelf components are cheap computers interconnected to form a Beowulf cluster. On the other hand, blade components are formed from computers that are mounted on a board that can be placed on a connector mounted on a rack (Englander, 2003). This process is through a much more similar way that peripheral cards are mounted onto a personal computer motherboard (Englander, 2003). Notably, blades are from standard off-the-shelf components or parts (Englander, 2003). Additionally, the COTS system contains a variety of hardware that originates from different vendors who also use different nodes; however, blade systems often have uniform hardware. Nonetheless, COTS is similar to blade since they are forms of cluster where there is interconnection between series of computers with intension of performing certain tasks. Clustered Computers Network Connectivity The network connection between computers in the cluster is never accessible from outside the cluster. This characteristic of network connection usually eliminates security concerns that may be required (Englander, 2003). However, authentication is necessary towards maintaining the integrity of the cluster. The Beowulf cluster has a single general front-end gateway server, which manages the computers within the cluster and provides outside world connectivity (Englander, 2003). Moreover, it provides a keyboard and monitor to be shared by all cluster nodes (Englander, 2003). In fact, in this case, each node is configured using its own operating system and hardware as well as own Beowulf clustering software. References Englander, I. (2003). The architecture of computer hardware and systems software: An information technology approach. New York: Wiley. Read More