Добавлен: 15.07.2023
Просмотров: 36
Скачиваний: 3
Non-traditional (or parallel) data processing is a way to perform a specialized class of tasks by dividing the task to be solved into separate subtasks and solving them on different nodes of the information system (or in different threads of the same node).
Unconventional data processing reduces the load on the nodes of the information system, transferring responsibility for the safety of data to each of the nodes of the system, thereby increasing the reliability of data processing.
Unconventional data processing may also involve storing data in a distributed system of node devices. In this case, the resulting array of nodes will be called the "version control system". A good example is GitHub: it is a decentralized system of distributed data storage (in this case, software code).
All data is stored in this system, even if it has been overwritten by the newly received data. the system will mark the existing data as outdated, and the received data as new. This way, the conversion history is based on all saved data, and you can track all changes in that data.
If parallelization of processing is organized on a separate node of the system, it can be implemented in several ways.
Instead of running large commands one after the other, run part of each command and store the intermediate result of the calculations for each command.
Another method can be used when storing data in parallel. in this case, one or more additional data drives are used, which store some of the data from the main drive. Thus, the read instructions will be read in parallel from different media.
To improve the efficiency of parallelization of calculations, the above methods can be combined. This significantly improves the overall data integrity of the entire node system, since in the event of failure of one or more nodes of the system, the remaining nodes can be used as temporary substitutes.
Building such a system also guarantees a strong complication of falsification of system data, since you will have to change data on absolutely all nodes of the system.
As a result, we can talk about the high stability of the system to falsification and discredit, its high security and ease of use by end users.