Exciting features of SQL Server 2012
Feature number 1 (Revolution):- Column store indexes
Feature number 2 (Evolution):- Sequence objects
Feature number 3 (Revolution):- Pagination
Feature number 4 (Revolution):- Contained database
Feature number 5 (Evolution):- Error handling
Feature number 6 (Evolution):- User defined roles
Feature number 7 (Evolution):- Windows server core support
Feature number 8 (Revolution):- Tabular Model (SSAS)
Feature number 9 (Revolution):- Power view
Feature number 10 (Revolution):- DQS Data quality services
Feature
number 1 (Revolution):- Column store indexes
Column store indexes are unexpected and awesome feature. When I read this
feature first time I was like, mouth wide open. You can get this feature when
you right click on the indexes folder as “Non-Clustered Column store Index” ,
as shown in the below figure.
So let’s quickly understand what exactly it does. Now Relational database
store data “row wise”. These rows are further stored in 8 KB page size.
For instance you can see in the below figure we have table with two columns
“Column1” and “Column2”. You can see how the data is stored in two pages i.e.
“page1” and “page2”. “Page1” has two rows and “page2” also has two rows. Now if
you want to fetch only “column1”, you have to pull records from two pages i.e.
“Page1” and “Page2”, see below for the visuals.
As we have to fetch data from two pages its bit performance intensive.
If somehow we can store data column wise we can avoid fetching data from
multiple pages. That’s what column store indexes do. When you create a column
store index it stores same column data in the same page. You can see from the
below visuals, we now need to fetch “column1” data only from one page rather
than querying multiple pages.
This feature is good to have and I personally feel it just mimics Oracle’s
sequence objects. Looks like it’s just a good to have feeling, if Oracle
has it why not SQL Server. A sequence object generates sequence of unique
numeric values as per specifications. Many developers would have now got a
thought, we have something similar like this called as “Identity” columns. But
the big difference is sequence object is independent of a table while identity
columns are attached to a table.
Below is a simple code to create a sequence object. You can see we have
created a sequence object called as “MySeq” with the following specification:-
- Starts
with value 1.
- Increments
with value 1 Minimum value it should start is with zero.
- Maximum it
will go to 100. No cycle defines that once it reaches 100 it will throw an
error.
- If you
want to restart it from 0 you should provide “cycle”.
- “cache 50”
specifies that till 50 the values are already incremented in to
cache to reduce IO. If you specify “no cache” it will make input output on
the disk.
create sequence MySeq as int
start with 1 -- Start with value 1
increment by 1-- Increment with value 1
minvalue 0 -- Minimum value to start is zero
maxvalue 100 -- Maximum it can go to 100
no cycle -- Do not go above 100
cache 50 -- Increment 50 values in memory rather than incrementing from
IO
To increment the value we need to call the below select statement. This is
one more big difference as compared to identity.In identity the values
increment when rows are added here we need to make an explicit call.
SELECT NEXT VALUE FOR dbo.MySequence AS seq_no;
There are instances when you want to display large result sets to the end
user. The best way to display large result set is to split them i.e.
apply pagination. So developers had their own hacky ways of achieving
pagination using “top”, “row_number” etc. But from SQL Server 2012 onwards we
can do pagination by using “OFFSET” and “FETCH’ commands.
For instance let’s says we have the following customer table which has 12
records. We would like to split the records in to 6 and 6.
So doing pagination is a two-step process: -
- First mark
the start of the row by using “OFFSET” command.
- Second
specify how many rows you want to fetch by using “FETCH” command.
You can see in the below code snippet we have used “OFFSET” to mark the
start of row from “0”position. A very important note order by clause is
compulsory for “OFFSET” command.
select * from
tblcustomer order by customercode
offset 0 rows – start from zero
In the below code snippet we have specified we want to fetch “6” rows from
the start “0”position specified in the “OFFSET”.
fetch next 6 rows only
Now if you run the above SQL you should see 6 rows.
To fetch the next 6 rows just change your “OFFSET” position. You can see in
the below code snippet I have modified the offset to 6. That means the row
start position will from “6”.
select * from
tblcustomer order by customercode
offset 6 rows
fetch next 6 rows only
The above code snippet displays the next “6” records , below is how the
output looks.
This is a great feature for people who have to go through pain of SQL Server
database migration again and again. One of the biggest pains in migrating
databases is user accounts. SQL Server user resides either in windows ADS
or at SQL Server level as SQL Server users. So when we migrate SQL Server
database from one server to other server these users have to be recreated
again. If you have lot’s of users you would need one dedicated person sitting
creating one’s for you.
So one of the requirements from easy migration perspective is to
create databases which are self-contained. In other words, can we have a
database with meta-data information, security information etc with in the
database itself. So that when we migrate the database, we migrate everything
with it. There’s where “Contained” database where introduced in SQL
Server 2012.
Creating contained database is a 3 step process: -
Step 1: - First thing is to enable contained database at
SQL Server instance level. You can do the same by right clicking on the SQL
Server instance and setting “Enabled Contained Database” to “true”.
You can achieve the same by using the below SQL statements as well.
sp_configure 'show advanced options',1
GO
RECONFIGURE WITH OVERRIDE
GO
sp_configure 'contained database authentication', 1
GO
RECONFIGURE WITH OVERRIDE
GO
Step 2 - The next step is to enable contained database at
database level. So when create a new database set “Containment type” to partial
as shown in the below figure.
You can also create database with “containment” set to “partial” using the
below SQL code.
CREATE DATABASE [MyDb]
CONTAINMENT = PARTIAL
ON PRIMARY
( NAME = N'My', FILENAME = N'C:\My.mdf')
LOG ON
( NAME = N'My_log', FILENAME =N'C:\My_log.ldf')
Step 3: - The final thing now is to test if “contained”
database fundamental is working or not. Now we want the user credentials to be
part of the database , so we need to create user as “SQL User with password”.
You can achieve the same by using the below script.
CREATE USER MyUser
WITH PASSWORD = 'pass@123';
GO
Now if you try to login with the user created, you get an error as shown in
the below figure. This proves that the user is not available at SQL Server
level.
Now click on options and specify the database name in “connect to database”
, you should be able to login , which proves that user is part of database and
not SQL Server
As a developer I am personally very comfortable with using “try/catch/throw”
syntax structure for error handling in c# or vb.net. Thanks to SQL Server team
in 2005 they brought in “try/catch” structure which is very much compatible the
way I as a developer was doing error handling in c#. It was nightmare
handling error using “IF” conditions and “@error” code before SQL Server 2005.
Below is a sample code which shows how “try/catch” code looks.
begin try
declare @n int = 0;
set @n = 1/0;
end try
begin catch
print('divide by zero');
RAISERROR ( ‘Divide by zero‘, 16, 1) ;
end catch
But what still is itching me in the above code is when it comes to
propagating errors back to the client I was missing the “THROW” command. We
still need to use “RAISEERROR” which does the job, but lacks lot of
capabilities which “THROW” has. For example to throw user defined messages you
need to make entry in to “sys.messages” table.
Below is how the code with “throw” looks like.
begin try
-- The code where error has occurred.
end try
begin catch
-- throw error to the client
Throw;
end catch
If you want to throw exception with a user defined message defined you can
use the below code. No entry need in the “sys.messages” table.
THROW 49903, 'User define exception.', 1
From SQL Server 2012 onwards use “Throw” rather than “raiseerror” , looking
at the features of “throw” looks like sooner or later “raiseerror” will be
deprecated . Below is a comparison table which explains the difference between
“throw” vs “raiseerror”.
|
|
Throw
|
RaiseError
|
|
User & system exception
|
Can generate only user exception.
|
Can generate user and system exception.
|
|
“Sys.Messages” table
|
You can supply adhoc text does not need an entry in
“Sys.Messages” table.
|
You need to make an entry in “Sys.Messages” table.
|
|
Original exception.
|
Original exception is propagated to the client.
|
Original exception is lost to the client.
|
In SQL Server 2008 R2 we had the ability to create roles at database level.
So you create customized roles at the database level and then assign them to
users. But at the server level or instance level we did not have options of
creating server roles. So if you right click on the “Server roles” you will not
find any options for adding new server roles.
Now that’s a serious limitation. Let’s say you have two sets of database
user one programmers and the other DBA’s. The programmers should be able to
fire insert, update and delete queries while DBA’s should be able to create
database, backup and do maintenance related activities. But DBA’s should not be
able to fire insert, update and delete queries. But now because you have fixed
roles the DBA’s get more access so they can even fire insert, update and delete
queries. In simple words we need flexible roles.
In SQL Server 2012 you can create your own role and define customized
permission for the role at a more granular level.
You can see in the below image how you can select permission at a finer
level and create customized roles which can be later assigned to a user.
This is a small evolution but an important one. Windows server core is one
of the flavors of Windows operating system. It is a GUI less version of windows
operating system. When you boot with windows core you would be surprised to get
a simple DOS command line as shown in the figure as compared to start program
files and crowded desktop short cuts. Because only necessary services are
enabled, we have less memory consumption, simplified management as many
features are not enabled and great stability. When we talk about SQL Server we
would love to run it over an operating system with minimal feature enabled. So
this is the most welcome feature and on production server using windows core is
definitely the way to go.
This is my personal top feature in SQL Server. Now the main goal of SSAS
(SQL Server analysis service) is to do analysis, i.e. Convert data in to
information. And SSAS achieves this by creating CUBES from data provided.
So the basic flow goes in 3 steps :-
- First data
is brought to central database (data ware house) using SSIS package. The
design of the data ware house system is normally in snow flake or star
schema, so that we can create CUBE’s effectively.
- Later
analysis services runs over the data ware house to create CUBES to give
multi-dimensional view of the data for better analysis.
- We can
then run different clients like EXCEL, SSRS etc to display data to
different sections of users.
Can you guess one big potential problem with the above 3 step approach?.
Give a PAUSE and think over it for a minute before you read ahead.
The biggest issue is simple business users CAN NOT CONTRIBUTE TO CUBES. I
mean if I am a business user who would like to take data from a excel sheet,
use my excel formula skills, derive conclusions and publish cubes, so how do I
go about it?. My personal belief is that the best business analysis can only be
done by business end users who actually do business on the field. They are the
best people who understand things and can create CUBES which are more useful
and logical.
Also if you notice the previous steps its highly technical:-
- Can a
simple business user create DB designs like snow flake / star schema?
- Can he use
the complicated SSAS user interface to publish cubes?.
- Does he
have the knowledge of using SQL Server analysis capability?
Note: - We will change our vocabulary so that we are
compatible with Microsoft vocabulary. We will term simple business users as
personal users hence forth.
Now personal users work most of the time with EXCEL and if we really want to
give analysis power to them, it should be inside excel itself. That’s what
power pivot does. Power pivot is plugin which sits inside EXCEL and gives
analytical capabilities to simple personal users to do analysis with data they
have in EXCEL.
Now EXCEL data is in tabular format with rows and columns. So if you want
publish this kind of analyzed data from EXCEL you need to have SSAS installed
in tabular mode.
So now if you compare personal users with professional BI the workflow will
be following:-
- Professional
BI personal will use SSIS, data flows, control flows etc.
- Personal
BI people can use import, copy past mechanism to get data in to EXCEL.
- Professional
BI person will uses SSAS , BI intelligence algorithm to do analysis. Once
analysis is done they will publish in multi-dimension format.
- Personal
BI people will use power pivot and excel formulas to come to an analysis.
Once analysis is done they will publish in tabular format.
At the end of the day both personal BI and SSAS will publish in a CUBE
format. So you can view the data from CUBE using SSRS , EXCEL or any other
mechanism.
So the personal BI user can use power pivot to do analysis. He can then save
the same as an simple EXCEL file.
You can then select import from power pivot, go to power pivot EXCEL file and
deploy the same in a tabular format.
To publish the same to tabular you can click on Build – Deploy tabular
project name.
Once deployed you should see the CUBE deployed in SSAS as shown in the below
figure.
Because the CUBE is created from tabular format we cannot use MDX to query
the CUBE. No worries, a new simple query language have been introduced called
as DAX (Data analysis expression). You can see in the below figure how I have
queried the “Sales 1” cube. DAX query starts with evaluate keyword, brackets
and then the cube name.
This article will not go in to DAX as our main concentration is SQL Server
2012 new features.
Every second project I have worked in my life always wanted a system where
in end users can go and create their own custom reports. Even though we have a
facility in SSRS for adhoc reporting it has huge limitations like you need to
install something on the client, works only with windows operating system and
internet explorer etc.
Power view is created for simple end user who would like to drag and drop
and create their own report using ad-hoc ways. It’s a simple Silverlight plugin
which gets downloaded and you get a screen something as shown below. End users
can now drag and drop the fields from right hand side, create a report and
publish it. Please note end users can not add fields that have to be added from
SSRS or Power pivot.
This feature would have been my top feature but due a serious limitation it
is not. “Power view only works with SharePoint”….I am sure you are feeling hurt
like me. Hope Microsoft makes this independent of share point.
If we visualize properly you can understand what the end GOAL of Microsoft is
to empower simple business users so that can do BI themselves. So a personal BI
user cannot get data in EXCEL, do analysis by using Power pivot and finally
create reports using the ad-hoc reporting tool power view.
This feature really touched by heart. When we talk about business
intelligence it’s all about DATA, DATA and DATA. One of the big problems with
data is that it can come in crude and unpolished formats. For instance if
someone has entered “IND” and you would like to change it to “India” so that
data is in a proper format.DQS helps you build a knowledge base for your data
and you can then use this knowledge base to do data cleaning. You can locate
DQS as shown in the below image.
Once you open DQS you will find three sections as shown below Knowledge base,
Data quality projects and Administration.
Knowledge base will help you define your validation rules. For instance you
can see in the below figure how we are creating a validation called as
“CustomerCode” and this validation checks if the data length is equal to
10.
You can also define correction rules like as shown one below. If you find
data as “IND” change it to “India”.
Once you have defined you knowledge, next step is to run this knowledge base
over a data. So create a DQS project and apply the knowledge base which you had
created as shown in the below figure.
You can then define where the data can come from and also you can map which
columns can have which validations. For instance you can see in the below
screen for country and customer we have mapped different domains. Domains are
nothing but validation rules.
Once done you can start the process and you would see a progress screen as
shown below of corrected values and suggested values depending.
Finally you can export the cleaned data to SQL Server, Excel or CSV.
