Peering database :-)
The idea of a network where everyone is both a client and a database server.
Suppose there is a database on a workstation. This database is part of a distributed database. The user forms a query to the database for sampling data from the columns CT1 and CT2, from the table TAB with the condition that the field CT3 = 5. A selection is made from the local database.
At the same time, from the list of available stations in the network included in this distributed database, several of the most profitable nodes are chosen, let's call them “peers”, by analogy with peer-to-peer networks. A copy of the request is sent to these nodes (data from the columns CT1 and CT2, where CT3 = 5, from the TAB table), indicating the request and the station generating the request (signature of the original source).
Each peer receiving the request conducts a sample of data in its local database (data from columns CT1 and CT2, where CT3 = 5, from the TAB table).
At the same time, each of them chooses his peers. And it asks if the station (peer) is ready to process this request from some original source. As a result, if the request has already been processed, a refusal is issued and another profitable recipient node is selected (in order not to resend the same request to the same peer).
Again, the selected recipients are sent a copy of the request (a sample of data from the column ..., where ..., from the table ...). And so it goes on and on, until ...
Until the limit is reached. The request specified limit: the "depth of the request" and the lifetime of the request.
Having completed the selection in the local database, the final nodes no longer send the request further, and the generated result is sent to the peer from which the request was received.
That, in turn, combines the result of the local sample and the results obtained from other peers and sends the processed data to the station that sent the request. And so on to the original source. Virtually each station generates a result based on local data and data from its “correspondents”.
Suppose there is a database on a workstation. This database is part of a distributed database. The user forms a query to the database for sampling data from the columns CT1 and CT2, from the table TAB with the condition that the field CT3 = 5. A selection is made from the local database.
At the same time, from the list of available stations in the network included in this distributed database, several of the most profitable nodes are chosen, let's call them “peers”, by analogy with peer-to-peer networks. A copy of the request is sent to these nodes (data from the columns CT1 and CT2, where CT3 = 5, from the TAB table), indicating the request and the station generating the request (signature of the original source).
Each peer receiving the request conducts a sample of data in its local database (data from columns CT1 and CT2, where CT3 = 5, from the TAB table).
At the same time, each of them chooses his peers. And it asks if the station (peer) is ready to process this request from some original source. As a result, if the request has already been processed, a refusal is issued and another profitable recipient node is selected (in order not to resend the same request to the same peer).
Again, the selected recipients are sent a copy of the request (a sample of data from the column ..., where ..., from the table ...). And so it goes on and on, until ...
Until the limit is reached. The request specified limit: the "depth of the request" and the lifetime of the request.
Having completed the selection in the local database, the final nodes no longer send the request further, and the generated result is sent to the peer from which the request was received.
That, in turn, combines the result of the local sample and the results obtained from other peers and sends the processed data to the station that sent the request. And so on to the original source. Virtually each station generates a result based on local data and data from its “correspondents”.
')
Source: https://habr.com/ru/post/17801/
All Articles