Victoria White
Distributed System Failure
December 16, 2013

There are two types of system structures that can be created. The first is a centralized system, which consists of one or more major hubs. All communication is processed through these hubs. This system setup provides security, to an extent, since all of the computing is done through a single computer. However, it also creates a single point of failure, if the main computer goes down the system is down. A distributed system is a collection of processors connected by a communication network. The processors may include microprocessors, workstations, minicomputers, and large computer systems. These processors are known by a few different names, sites, hosts, nodes, computers, and machines. There are a couple major reasons for creating a distributed system, these reasons include resource sharing, communication, reliability, and computation speedup. However, there are a few failures that may occur with a distributed system these failures include link failure, host failure, storage media failure, and scalability.
The first failure, link failure, occurs when the connection between two parts of the system fails. When this takes type of failure takes place the two parts of the system connecting can no longer communicate with each other. To detect link failure, a procedure known as handshaking is done. With this procedure first the host that is still functioning will continue to send I-am-up messages to the other host. After a certain period of time if the sending host does not receive a response from the receiving host it can assume the connection to the receiving host has failed. If the sender does not receive a response the next step can be one of two things. One, wait another designated period of time to receive a response, or two an Are-you-up message can be sent to the host in which the connection may down to. If this direct link from host A to B has failed the failure must be broadcasted to every host in the system. The broadcasting of this information is necessary so that the routing tables can be updated to skip over this line of communication.
Secondly, host failure can occur, this happens when the host becomes isolated from the system or there is a type of hardware failure that has taken place. In both scenarios of host failure communication to this host is lost. To determine if host failure has taken place, the handshaking procedure can be done. This procedure is done in the same way it is done when detecting link failure, a message is sent then a time period is waiting for a response. If the response has not been received one of two messages can be sent again and if a response is not received again it can assume the host is down. If the system believes a site has failure, and can no longer be reached by another site, then all sites must be notified. Sites must be notified so that they no longer attempt to use the services of the failed site. If the site served as a “central coordinator” and was the main site for a specific service, a new site must be elected to take over. However, if the site has only isolated itself and is still up and running but not communicating, then a situation in which there are now two “central coordinators” for one service. This type of situation can create conflicting actions since they will be trying to complete some of the same processes. Another type of failure that can occur is a storage medium failure. This type of failure happens when a device where data can be placed, kept, and retrieved from goes down. If storage medium failure occurs there is a loss of productivity in the system. This is due to the fact that some data is no longer able to be accessed in order for the processes needing the information to be completed. Storage medium failure can be solved with a RAID (redundant array of independent disks) setup. RAID will allow for a redundant solution since all of the disks are independent from one another, if one of the disks does down there are still more where the information is still available. Storage medium failure is a common type of failure since it can be caused by a simple hardware failure. Hardware failure is a failure that can cause any type of system whether it be a simple home system, a centralized system, or a distributed system. Scalability is also a common issue that can occur in any type of system. Scalability is the capability of a system to adapt to an increasing service load placed on the system. When a distributed site is created it is set up to handle a certain amount of users, services, and locations. If the number of users using the system increases, which in turn means a larger number of processes to be completed, all of the memory available on the system may be used. If this occurs the system will lose productivity since it will slow response time. When the system is upgrading or the amount of information needing to be handled increases a way to make sure the system can handle it is to add resources. Although adding more resources to the system may help handle the greater flow of information, it could also create other possible issues and the cost of increasing the system can be high. There are some failures with distributed systems that can also occur with centralized systems. Two of the failures are host failure and storage media failure. These two failures can occur in both types of system setups because they are basic failures. Host failure can occur because hosts may lose communication between one another in any type of situation for numerous different reasons. Storage media failure is also common since it can be caused by a simple hardware failure. Storage media failure can create numerous issues for the system unless a system such as RAID is put in place to ensure information is still accessible if one of the storage devices happens to go down.
Failures can occur in any system that is used whether it be a distributed, centralized, or a different system. Common failures that can occur with a distributed system are link failure, host failure, storage media failure, and scalability. There are solutions to fix the failures that may take place once they are detected.

References:
Silberschatz, A., Baer Galvin, P., & Gange, G. (2012). Operating System Concepts (8th ed.). Retrieved from The University of Phoenix eBook Collection database.