Database Reference
In-Depth Information
◾
Design queries that can be distributed across multiple federation members. hen, combine
the query results together.
◾
For better performance, run distributed queries at once on multiple federation members. For
testing and for simplicity, run the distributed queries sequentially.
◾
After the results of a distributed query are returned, the results are reassembled and the reas-
sembled results are returned to the application that issues the query. For testing, a partial
result from each of the federation members can be returned.
◾
Extract data from multiple federation members into a standard database before executing a
function or stored procedure.
8.4.1.3 Data Distribution across Multiple Federation Members
To decide how to distribute database objects to multiple federation members, consider the follow-
ing strategies:
◾
Store all the global data in separate nonfederated databases.
◾
After an atomic unit has been stored in a federation member, its federation key can be stored
in a nonfederated database for later lookup.
◾
If sequential big numbers are used as federation keys, distribute the atomic units across
based on prespeciied ranges of federation keys. he ranges can be deined as subsets of
sequential big numbers.
◾
If GUIDs are used as federation keys, use a hashing algorithm to convert the GUIDs
to numerical numbers and distribute atomic units according the range of the numerical
numbers.
◾
To distribute atomic units evenly and automatically, one may consider distributing those
units in a rotational manner. he diiculty of the method is that it will not know where to
locate a speciic atomic unit unless the same atomic unit is distributed to every federation
member in the same federation.
◾
For better performance and easier management, an atomic unit can be distributed to each of
the federation members. Newly inserted data can then be written to the atomic unit hosted
by each of the federation members. In such a way, multiple applications are able to access the
atomic unit simultaneously through diferent federation members.
8.4.1.4 Applications and Federation Logic
Applications access federations through federation logic. he following are some design strategies
related to applications and federation logic implementation:
◾
Use programming units to maintain relationships and import/export data across multiple
federation members.
◾
Use a single user account to create multiple federation members, which share the same secu-
rity credentials.
◾
To allow an application to connect to multiple federation members simultaneous, a connec-
tion management pool should be created in the application. In SQL database, ADO.NET is
the technique for carrying out connection management pooling.
◾
Applications supported by a federation should not be designed to rely on data transactions
between two federation members. he applications should not join two database objects
