Training Notes on 2nd Day of SQL Analysis Sever

Realtime Warehouse - Do we really have such thing ?You can't analyse the data when it is not consistent for some time

Find the difference between - ER Model Vs Dimensional Model

Dimensional Model rules
1) Almost Dimensions Tables are mostly refered 1 to many to Fact Table
2) Summerizable numerical values are Measures
3) Dimension is an indipendent entity of a specific concept. For example Jan, Feb, Mar, etc are Time dimension and India, USA, Russia, etc are location dimension blah blah
4) Every thing that you see in world are grouped
5) From one Dimension one or more hierarchy is created
6) Think for only ANALYSIS, but not for information that is hidden and unwanted. The prime target of this tool is only to understand the progress and analyse the business but not for getting the details of the transacion
7) Remember that, how better you can give the user the Drill down capability

Customers - Dimension

ID CustName City State Country MaritalStatus Gender Tel eMail

How many hirarchies you can create out of the above individual labels?

H1) GenderwiseMaritalstatusTelnumber - Wrong Hierarchy
H2) H1+eMail - Wrong Hierarchy
H3) StateWiseGenderMaritalStatus

Note: Hierarchies are created only the grouping possible labels, where as Tel and eMail can't be grouped, using them in hierarchy is a bad design

Types of Hierachies
1) Balanced - All leafs will have equal number of Members / LevelsRule 1 - All leave memebers should be at the same levelRule 2 - The parent member don't have the values, they are procured from the childs

2) UnBalanced - Not necessarily you will have the equal number of levels or members

3) Ragged - Ragged are more similar like Balanced but just missing some parents
Never Make the OLTP Primary keys same PKs of Application System
Dimension tables are almost smallFact Tables are generally huuuuuuuuuuuuge

Schemas
1) Snow-Flake - Bad Design
2) Star

Thing to remember: If report is taking long time to conceptualise, it is inotherwords conveying that the database design is bad. So, if the report is taking less time to conceptualise on the DB what you have, is a good design as, visualised the report while designing
"Change the granularity is beyond the scope"

Dimension are not required to create repeatedlyRole Playing Dimension - Similar to the Table Alias

Measures are of 3 types
1) Additives
2) Semi Additive
3) Non Additive

Slowly Changing Dimension

-----------------------
If you have something to share in person on this post, pl drop me a mail at dskcheck@gmail.com with the title in the subject.

Comments

Popular posts from this blog

Network Intrusion Detection using Supervised ML technique

Common mistakes by Interviewer

Keep the system active, to avoid the auto lock