1. Synchonization in D.S.
1. Synchonization in D.S.
UNIVERSIDAD CATOLICA DE COLOMBIA
 
1. Synchonization in D.S.
1.1 Algorithm of Lamport
2. Other algorithms
3. Mutual Exclusion
4. Algorithms Of Election
5. Atomic Transactions
6. Distribuited detection of locks
Spanish Summary
SYNCHRONIZATION OF CLOCKS
A distributed system must allow the appropriate use of the resources, must take charge of a good performance(discharge) and of the consistency of the information, beside keeping all these operations sure.

The process synchronization in the distributed systems turns out to be more complex than in the centralized ones, due to the fact that the information and the processing is kept in different nodes. A distributed system must support partial and consistent conference of all the cooperative processes and of calculation. Such conference can be provided by the mechanisms of synchronization.

The word synchronization is defined as the way of forcing a partial or total order in any set of events, and is used to refer to three problems different but related between(among) yes:

1. The synchronization between(among) the issuer and the recipient.

2. The specification and control of the common activity between(among) cooperative processes.

3. The serialización of competing accesses to objects shared by multiple processes.

Referring to the methods used in a centralized system, which uses semaphores and monitors; in a distributed system algorithms are in use distributed to synchronize the common work between(among) the processes and these algorithms have the following properties:

· As a characteristic of the distributed systems, it is unacceptable that centers in a node, to all that information that is relevant for the capture of decisions that have to carry out the processes, since it is common that they do it being based on the information stored locally.

· it is necessary to to meditate or foreseeing the possible points of failure(judgement) of the system, since in a distributed system the flow of work has not to turn affected if the process is interrupted in some node.

· Due to the fact that there can exist a great number of nodes of processing, there does not exist a clock or source(fountain) of common time for all of them. This is simplified to the fact of in that time happens an event with respect of other one.
The word 'clock' is in use for referring to the device with which they count(tell) almost all the computers, and that it(he,she) takes charge registering the time. There turns out to be more appropriate the use of the word 'chronometer'. The chronometer of a computer consists of a crystal of quartz that, if he is had subject to tension, ranges with a definite well frequency. The chronometer has a record associate book-keeper(meter) and another president. For every oscillation, the book-keeper(meter) is decrementado in 1 and when this one comes to 0, there is generated an interruption (so called brand(mark) of clock) and returns to load with the value stored in the president. Of this form it is possible to programme to the container with a value of 60 oscillations, which they would represent a second.

In a centralized system, there do not manage to exist problems caused by the fault of the chronometer, because the processes refer to the same chronometer. In a distributed system this changes, since every node possesses(relies on) his(her,your) own(proper) chronometer and it turns out impossible to guarantee that his(her,your) crystals will range with the same frequency. In a distributed system number of computers exists n and them n crystals will range to a bit different rates, which provokes a loss of synchrony and that on having read them, have different values. This difference is called a distortion of clock. Since(as) consequence of this distortion inconsistencies might be generated, since the programs wait that the time associated with a file, process or message is correct and independent from the site(place) where it was generated.

For the majority of the cases, it(he,she) is enough that all the computers have registered the same hour, though this one is not the royal(real) one. To this it is possible to be call him a logical clock. There exist also physical clocks, which are external to the computers and take charge taking a count of the real time. This type of clocks are useful for some systems, especially for those of real time. The physical clocks are recipients that catch signs of radio(radius) of short wave or signs of satellite, from institutions dedicated to taking a record of the almost exact time.

There is done emphasis in which the communication and the synchronization are necessary to reach the correct cooperation between(among) processes. The previous thing stems on having affirmed that some forms of communication need a previous synchronization.

The cooperative processes have to to be able to inform between(among) them, that they have completed certain action(share). From the point of view of the distributed systems, the term(end) synchronization refers to the temporary order of the set of events produced by competing processes.