字体： 小 中 大 特

by nicetomeetyou (老兵新传) at 2007.10.24 10:13 (#4012350@0)

Backend:

(1) Finish data warehouse design and implementation.

(2) Figure out data sources, and data targets, their locations, the drivers to load them. There are various types of data source: Text, Excel worksheet, other database, Web service, XML stream, etc.

(3) Make up strategy plan, scheduling, performance threshold, monitoring tools.

(4) Assign roles to operator, designer, manager, and admin.

(5) Create metadata for data source, target, intermediate SQL gates, customized transformation, input/output links. Re-use those metadata from repository if necessary.

(6) Make suitable configuration so that we get report log in appropriate level.

(7) Start trial run. Figure out the performance bottleneck. Most of time in some performance-critical cases, you must make some native transformation gates to replace system offered standard ones, using database-specific SQL to create high performance stored procedure / user functions. Based on dependency analysis, you may have to use parallel mode to load and transform data.

(8) Implement on production after full-fledged QA Test.

Front-end:

Thin Client n-tier model / Thick client / server model. Access / Role Management. Design / Implement report interface. Pass user acceptance phase.