Phase I -01. Towering building, foundation primary -1-3 large site architecture evolution

web2.0 era of

the early development of traditional web

server and database file


server sites only contain other data, such as user uploaded picture, there will be an additional file server storage.
Database is also a separate server.


User multiplied exponentially increases, the amount of the user's query. All the pressure will be on the database, the introduction of caching middleware. Most users will fall query caching middleware inside, rather than straight into the database. To ensure that users experience enhanced response speed will be accelerated.

Single node into the bottleneck.


Most of the request is to read the database, the database is to write a small part. Separate read and write the database. Library to synchronize the timing of the main data from the database,

the database do not carry the sub-library sub-table. There are multiple databases with a table of data according to certain rules and algorithms, hash in a different database, a practice also known as distributed database. This is the last resort we split the database. Only when the database size is very powerful, it will consider using it. Usually a single database table to reach 700 million or 8 million, we will consider doing so. Because the performance of our database would drop dramatically,

once sub-library sub-table, and distributed the primary key, which is the primary key for each piece of data within each of us a table, and then not be able to use a self grew. We must use the distributed primary key, globally unique primary key.

Solr introduction or elaticSearch search engine.


Split service. Distributed transaction, the user's request may arrive at the same time a number of subsystems, so the problem of distributed transactions also must be considered.







 

End