Distributed calculation

The calculation distributed or distributed , so known under the name of system distributed , consists in distributing a calculation or a system on several distinct Ordinateur S. Distributed calculation is a parallel form of Calcul, on the other hand it differs from the Grappes of waiters ( cluster computing ), in what the Ordinateur S which carry out calculation are not typically dedicated to distributed calculation and always does not have same the materials and filing system, whereas the bunches ( clusters ) include/understand most of the time material specific and dedicated to a precise task.

Distributed system

A current definition is that a distributed system is an autonomous collection of stations which are connected using a communication network. Each station carries out components and uses a Intergiciel, which deals with activating the components and coordinating their activities of such kind that a user perceives the system like a single integrated system.

This definition implies an important property of the distributed systems, that is to say that the distribution is generally hidden for the user and the application programmers. He prefers to see the system like only one and single system and thus to hide the complexity of the distribution as much as possible and to increase the transparency distributed system. That makes it possible to develop as much as possible the applications in the same way that the Système centralized. A distributed system is generally separable in several entirely autonomous components. There does not exist component Master which manages the others and each one is thus responsible for its own operation. That allows, inter alia, to have a heterogeneity in the technology used for each component, they can be written in various computer programming languages (Java, Cobol, C++, etc) and be carried out on various filing systems (Eg Linux, Windows, etc). The autonomy of the components makes that the systems are carried out simultaneously (concurrent Programmation). Moreover, contrary to the centralized system, the distributed systems have several points of failures (problem of components, network, traffics, etc).

Requirements of the distributed systems

The need to use a distributed system is often derived from Exigence S not functional calculuses is:

Extensibility (" scalability") - the distributed systems easily allow an expansion so necessary.
Ouverture - the components of the distributed systems have well defined interfaces what enables them to be easily extensible and modifiable. The services Web are an example of distributed system which has a large opening.
Heterogeneity - the components can be written in various languages on various machines. For example, the elements of a waiter can be programmed in C++ and the customer in Java.
Access to the resources and division - the distributed systems provide a means of sharing the resources is, the hardware, the software and the data.
Fault-tolerance - the distributed systems make it possible to be more fault-tolerant than the centralized systems, because it makes it possible to retort the components easily.

Projects

The project “pioneer”, in the use of distributed calculation is the SETI@Home the purpose of which is to try “to listen to” space, in collaboration with SETI ( Search for Extraterrestrial Intelligence ) in order to detect a possible trace of extraterestre activity in space.

But the need growing for data-processing computing power in the medical research and other fields is especially what caused the increasingly important use of this technology. The cost of the supercomputers being too high, it is thus planned to use the computing power “available” of computers at rest (or under-utilized). The process often consists of the installation of a software which downloads raw data starting from a waiter, works over again them (“draft”) in a transparent way for the user (by using only the computing power not used by other applications), then returns the results to the waiters.

Certain systems of distributed calculation are opened with the public via Internet (external cf resources) and attract the users easily. The competitive spirit, the passion of the numbers, are filled by the provision of very detailed statistics. One will be able to thus see the evolution of the project since his beginnings, the personal classification of the user compared with the other machines, the computing time carried out by the owners of a given architecture, or by such or such operating system (Linux, Windows, Macintosh…).

Distributed calculation is also an active topic of research, with an abundant literature. Most known of the conferences on distributed calculation are “The International Conference one Dependable Systems and Networks” and “ACM Symposium one Principles off Distributed Computing”. There exists also the review “Newspaper off Parallel and Distributed Computing” ().

A big part of the human genome thus could be deciphered by the Net surfers of the program “Decrypthon”. Projects also in progress, are centralized by the World Community Grid aiming to the analysis of proteins and the development of solutions against the AIDS.

Let us note however that some affirm that the use of this technology allows the multiplication of patents on drugs and information which would be more effective in the public domain. Décrypthon for example contributed to the decoding of the human genome, including one enormous part is patented by the drug companies.

External resources

the calculation distributed on Internet
Course on the systems distributed

Projects of shared calculation:

World Community Grid : http://www.worldcommunitygrid.org/ is supported by IBM and point of disjunction of many project such as Rosetta (calculation of proteins) and well of others…
Distributed.net: http://distributed.net/ lodges many projects, of which one is the research of the optimal rules of Golomb. Some will say that Distributed.net is not any more one project with nonlucrative goal since project principal RC5-72 brings a reward of laboratories RSA http://www.rsasecurity.com/rsalabs/node.asp?id=2100;
SETI @home : http://setiathome.berkeley.edu/ seeks signs of an extraterrestrial intelligence;
GIMPS : http://www.mersenne.org/ to seek the prime numbers of Mersenne. It would seem that there too one cannot qualify the project of project to nonlucrative goal since he considers a reward of the http://www.eff.org/ Electronic Frontier Foundation;
United Devices : http://www.ud.com/ the greatest sales network of distributed calculation;
Genome@Home: http://www.stanford.edu/group/pandegroup/genome/ Sequencing of Protein S in the human Genome, this project is completed since the end of the sequencing of the human genome was announced on April 14th, 2003. The university of Stanford replaced it by Folding@Home;
Folding@Home : http://folding.stanford.edu/ simulations in the way in which the Protéine S are folded up and are unfolded on themselves;
Decrypthon: http://www.decrypthon.fr allowing to carry out calculations of comparison of the Protein S which made its return in March 2005;
Proteins@home New Project emanating of the Polytechnic school (France) on a model of Prediction of structure of proteins:
FightAIDS@Home : http://fightaidsathome.scripps.edu/ Project to fight against the AIDS;
Seventeen gold Bust: http://www.seventeenorbust.com/ Project to solve the problem of Sierpinski;
Africa@Home ; http://www.africaathome.org/ Project of fight against the Paludism.

BOINC:

Boinc is a platform of shared calculation making it possible to make function several projects on only one interface:
Official site of BOINC

See too

Networks P2P
distributed Intelligence

Random links: