First, the classification table
1, the entity table
Entity table , generally refers to the existence of a real business objects, such as users, commodities, businesses, salespeople and so on.
2, dimension table
Dimension tables , generally refers to a corresponding number of operational state, number interpretation table. It can also be called code table.
Areas such as tables, order status, payment, approval status, Categories and so on.
3, transactional fact table
Transactional fact table , generally refers to data generated as the business continues to occur. Characterized by the event will not change .
Generally, for example, transaction flow, operation log, a library storage records and so on.
4, cyclical fact table
Cyclical fact table , generally refers to data occur as the business continues to generate changes (updates, new) is.
The difference is that with the transactional data will advance as the business cyclical changes.
Such as orders, order status will change periodically. As another example, leave, loan applications, with the approval of the state in cyclical changes.
Second, synchronization strategy
The type of data synchronization strategy include: full-scale, incremental table, add and change table, table zipper
- Full Scale : data storage complete.
- Incremental Table : newly added data storage.
- Add and change table : data storage and changing new additions.
- Zipper Table : New and changes made to the table on a regular basis merger .
1, synchronized strategy entity table
Entity table: for example, users, products, business, salespeople, etc.
Entity table data than smaller: the whole amount can usually be done daily, every day is to keep a complete data . That is the whole daily amount.
2, dimension table synchronization strategy
Table dimensions: for example, order status, approval status, Categories
Smaller than the dimension table data: full daily amount can usually do is keep a complete data every day. That is the whole daily amount.
Description:
1) Data for the status may change can be stored daily full amount .
2) did not change the dimensions of the objective world (such as gender, regional, ethnic, and political components, shoe size) can store only one copy of a fixed value.
3, transactional fact table synchronization strategy
Transactional fact table: for example, transaction flow, operation log, a library storage records.
Because the data does not change, but also a huge amount of data, so you can only sync new data every day, so you can make daily incremental table that day to create a partition storage.
4, cyclical fact table synchronization strategy
Cyclicals fact table: for example, order, leave, loan applications, etc.
This type of watch from the perspective of the amount of data, save a full daily amount, then the amount of data is too large, too much redundancy. If the data does not reflect changes in daily increments of words.
New daily and the amount of change , including the day of the new and modified. Generally the table, most of the day is sufficient calculated data. But this is still not resolved able to get a certain historical point in time (time slices) of the slice data.
So they need to take advantage of the new daily and change table, make a list zipper for easy to take a snapshot of a time slice data. So we need to get new and daily variation.
Third, the paradigm
Relational database design, in accordance with certain specifications, aims to reduce data redundancy,
The industry paradigm are: first normal form (1NF), second normal form (2NF), Third Normal Form (3NF) , Bath - Cape Cod Paradigm (BCNF), fourth normal form (4NF), Fifth Normal (5NF).
Paradigm can be understood as a data table design table structure, in line with the standard level.
Use paradigm fundamental purpose is:
1) reduce data redundancy, as far as possible so that each data only once.
2) to ensure data consistency
The disadvantage is that acquiring data, required by the final Join splicing data.
Fourth, functional dependency
Five, three paradigms distinguish
First Normal Form
Second Normal Form
Third Normal Form
Sixth, the relationship between modeling and dimensional modeling
Relationship Modeling
OLTP relational model is mainly used in the system in order to ensure data consistency and to avoid redundancy , business systems division table are followed in third normal form
Dimensional modeling
OLAP dimensional model is mainly used in the system, because the relational model, although less redundancy, but in large-scale data, cross-table analysis of statistical query process, will result in multi-table association, which would greatly reduce the efficiency
Therefore, the relevant various tables organized into two types: the fact and dimension tables two kinds. All around the dimension table to explain the fact table.
Online analytical processing (On-Line Analytical Processing, OLAP)
Online Transaction Processing System (On-Line Transaction Processing; OLTP)
Seven, Snowflake model, star model, Constellation model
On the basis of dimensional modeling of the divided into three models: the star model, snowflake model, Constellation model.
