Four Common Failures in a Distributed System

Four Common Failures in a Distributed System

POS/355

January, 14th, 2013

Distributed System Overview

A distributed system is a collection of dummy computers connected to a network of distributed middleware. This allows the computers to communicate to each other and also share resources. While allowing the end user to use the dummy computer as he or she would use a single integrated computing facility (Emmerich, 1997). A huge benefit of using a distributed system is the reliability factor that exists, if there is a failure that occurs like a site failure for instance, then the remaining sites can continue to operate. A huge advantage of a distributed system is the system's capacity to send messages over very long distances making collaboration on projects, whatever they may be, much easier.

There are various strategies in the routing scheme that can help reduce the impact of system failures. Fixed, Virtual, and Dynamic Routing can be employed within a distributed system. In fixed routing (as the name indicates) a path is specified in advance and will not be changed, in Virtual routing scheme's a path is only fixed for the duration of one session and with Dynamic routing the path is chosen when a message is sent. Each schema has its strengths and weaknesses but at the end of the day, in any of these common routing schemes, failures can and will occur. Understanding more about the system architecture whether it be centralized or more specifically, distributed, can help us realize exactly how an error or failure might occur.

Client/Server Cluster Architecture

Interprocess communication is a fundamental mechanism in a distributed system; the cluster is the entire group of computers functioning as one giant resource so a messaging system is crucial to the overall operability of the distributed system. In a client/server cluster architecture you have two sides; one being the client side or the individual workstation and the server side. From the bottom up and on both sides you have the hardware platform that both operate/use, one layer up, we will see the client/server OS, then the communications software on both sides and the DBMS on the sever side as well if there is a database application layer to the system. At this tier there is protocol interaction that takes place between the client and the server.

One level up from that we have database logic in which the client workstation sends requests and the server responds. The client side has to more levels above this that encompass the application logic and the presentation services. In a more generic architecture the application logic can be on both the client and server sides which functions as