Can applications query datawarehouse repository directly; or should they always access the data via data marts? -
i plan on implementing inmon type of datawarehouse solution small/medium size company. datawarehouse solution have 3rd normal form repository , set of data marts. data flow "online" datasources (oltps) data warehouse repository , data marts. believe have grasp of theory. i've read few books ("the data warehouse mentor" laberge, books kimball , inmon) have questions regarding real-life solutions , best practices.
questions:
- is idea have applications (reporting systems etc.) execute queries against data warehouse repository (i'm referring central 3nf data storage), or should access data through data marts?
- is there standard naming convention data repository , data mart objects (schemas, tables, fields etc.)?
- can point me examples of real-life datawarehouse schema examples? i've reviewed mssql adventureworksdw.
i'd appreciate feedback. thanks.
1) depends. shouldn't reason have dimensional data marts @ if 3nf performs. data marts typically there aggregation & performance reasons. brings question of "why have 3nf layer if queries happen in data marts anyway?"
2) use best practices company has defined. there sorts of standards across industry, use have already.
3) real-life inmon-style data warehouses exclusively exist @ large companies can afford immense time , political build immense data dictionary , "corporate information factory" inmon defines, or small companies have few source systems read data from. if don't integrate every system, comes online, there problems in future when discover 3nf model doesn't fit business exactly. kimball data marts can built starting 1 business area , expanded on time include others, rather having front.
Comments
Post a Comment