Most database systems are now able to achieve the constant time to return a certain amount of data. For example, Oracle databases, according to primary key data in the one million data were obtained from 10, with 100 million data in 10 data were obtained from the difference in time not much.
Avoid a large amount of data to take a database, taking large amounts of data must use paging.
This is basically a database system design is now a lot of the basic code of practice. ERP system, more than one million data tables many to many tables in any one of a collection of all database servers will lead to prolonged stagnation, and the impact of other online users of the system response speed.
In general, daily operations, the situation shown in the following pages, each between 1-100 to obtain the data, the system response speed is fast enough, the client basically not a particularly long pause. This is an ideal design. This is also the large-scale database systems are often used ODBC, ADO, and so common database connection components rather than a specific, higher-speed dedicated database connection components reasons. Because the system bottleneck is the database (Database) aspects (large amount of data), rather than the client (client-side take only a small amount of data).
In the B / S database systems, paging very common. Early database systems often have a client program to take large amounts of data to do one-time buffer. This is no longer the special needs of the two main reasons:
Database itself cushioning technology greatly enhanced.
Most databases are commonly used in automatic data automatically placed in memory buffers, to improve performance.
Database connection component buffer technology is also improving.
Including ADO database connectivity components, including some of the data will automatically result set (result set) to the buffer, and the results are good. Relatively new database connectivity components, such as Hibernate has joined some of the data result set buffering function. Of course, there are some components of the database connection is not the result sets of data buffering, such as JDBC Driver, but within a few years the situation should have improved. Are also some less successful data buffer, such as EJB entity in Bean, the performance is not satisfactory, the Entity Bean data is also placed in memory, perhaps because too much memory for the sake of.
Comparatively speaking, today's programmers to write client-side data buffer, the buffer can be more than the above two effects have been more difficult.
|
