DC-projects distribute their work broken down into individual work packages, so-called "work units" (WUs). These work units are distributed to the available computers, processed by them and finally sent back to the server. The duration of a work unit varies from a few minutes to several months, depending on the project. Often there is also a (deadline), within which the WU must be finished.
Distribution of work units
For the most projects, the client periodically contacts a central server and loads the next work units via the http protocol, similar to a web browser.
Smaller projects in particular, which do not attract so many participants, often can not afford the expense of an automated centralized administration of the work units. That's why you can reserve work units for some projects (especially in the mathematics) by sending an e-mail to the operator of the project specifying which area you want to edit yourself. More and more frequently, participants also coordinate on entries of the areas to be reserved in the forum of the project.
Multiple assignment of work units
Since not all participants in a project stay until the end, many project work units do not receive any results. Work units must therefore be issued multiple times until all results are returned.
Unfortunately, with increasing popularity of distributed computing, people have also been involved in projects for which their personal rank in the statistics is more important than the outcome of the project. Then, as with SETI@home Classic happend clients were manipulated so that falsified results (in a very short time) were returned, the redundancy was above all, which made it necessary to compare several results to exclude individual WU manipulations.
Last but not least, computers are not as reliable as they often seem to us. Especially with very lengthy calculations, it can happen that a bit in the memory tilts due to minimal defects in the hardware or in other ways wrong results come about. Even the use of processors from different manufacturers can lead to different results. The multiple assignment of work units, the so-called quorum, protects against these rather random errors on individual computers.
Projects without work units
Although it may sound strange, there are indeed projects that work without the distribution of work units. For some cryptography projects, such as ECCp-109, it is sufficient if the participating computers use random initial values for their calculations, so that the central creation and management of the work packages is not necessary. It is then not important to calculate a certain number of work packages, but simply so many, until enough results with certain properties have been calculated and the encryption can be cracked.