Pos355 Distributed System Failures

Four Types of Failures for a Distributed System

There are two types of system architectures that can be used when setting up a network. They can either use a distributed system or a centralized system. Both systems can have failures.

Distributed Systems

A distributed system is a collection of loosely coupled processors interconnected by a communications network and running distributed middleware components. Distributed systems can vary in size and functionality, they may be co-located or across great distances. The computers can communicate to each other without intervention and share resources such as files, databases, hardware and software. They establish one computing environment and share characteristics such as high availability, fault tolerance, and scalability.

Types of Failures

Halting failures

A component simply stops working. There is no way to detect the failure except by timeout: it either stops sending an "I'm alive" (heartbeat) message or fails to respond to requests. The result is that your computer freezes and becomes inoperable.

Fail-stop or Byzantine failures

This is a halting failure with some kind of notification to other components. A database server telling its clients it is about to go down is a fail-stop.

Network failures

This is a break in the network’s connectivity, it could be due to physical or logical (software) failure at the network layer.

Omission failures

These include a failure to send or receive messages primarily due to lack of buffer space. This can cause a message to be discarded with no notification to either the sender or the receiver. This can happen when components become overloaded.

Centralized Systems

Some of these failures are not limited to happening in just a distributed system, they can also occur in a centralized system. There are a few types of centralized systems.

1. Computers that operate as a single device; do not need to interact with other devices to work.

2....