RSS

Monthly Archives: April 2015

T-SQL Tuesday #065 – Slowly Changing Dimensions

tsql2sday150x150_thumb_2aa4ea0f I’ve been focusing a lot of my study time on data warehousing lately. I’ve been supporting the system and storage of data warehouses for a while but lately have been digging into the developer topics.

What I learned over the weekend is how to build a working, slowly changing dimension in SSDT. Thanks for the challenge #tsql2sday and @SQLMD!

 

The Problem

Dimensions are the tables we design to make data look good in a pivot chart. They are the tables that describe our facts. Customer is a good example of something that could be a dimension table. For my challenge I decided to use virtual machine as my dimension.

The problem is, what if a VM’s attributes change? 4 cores, that was yesterday.. today PRDSQLX has 24 cores. What if someone deletes a VM, how many cores did it have?

I can get the current status of my VMs by using the source system, but the problem is the history. I can pull a snapshot of what VMs I have in my environment every day from the source system. I could just make copies of that data and slap a “PollDate” column on the table. Viola, I have everything I need, and about 1000x more than I need.

There is the problem, how do I collect and save a history of my VM’s attributes?

Solution

Each column in my VM table can be of 3 basic types http://en.wikipedia.org/wiki/Dimension_table

Type 1. Simply overwrite this value… it changes a lot and I don’t care about history (eg. what host is the VM running on)
Type 2. add a new row to maintain history… if one column in my VM row changes, I get a whole new record in my dimension
Type 3. add a new column to keep a limited amount of history… add some columns like previous_num_cpus and previous_previous_num_cpus and move data to that as it changes

So we have to take the data we get on a nightly snapshot of the source, and compare it to what we have in the destination, then do a conditional split. I’m sticking to handling these differences:

New VM – insert with NULL validto (easy)
Deleted VM – change validto column (create staging table and do an except query)
Change in Type 1 Col – update existing VM row with NULL validto column, (easy)
Change in Type 2 Col – insert new row with NULL validto column, change previous record’s validto date (a little tricky)

That logical split can be made easier by using the Slowly Changing Dimension task in SSDT. It pops up a wizard to help you along the way and completely set you up for several failures which I am going to let you learn on your own :]

Step 1. Setup an initial loading package.

This will make it handy to restart your development.

Query the source in a data flow OLE DB Source
Tack on a few extra columns, validfrom, validto, isdeleted, sourcesystemid in the SQL command
create the destination table using the new button ( this is pretty handy to avoid manually lining up all datatypes )
use the new button again to create a dimVM_staging table for later
Add the task at the beginning of the control flow to truncate destination or dimVM table
Run the package and be careful not to accidentally run it since it has a truncate

Step 2. Create this monstrosity

Control Flow
dimVM_scd_control_flow

Data Flow
dimVM_scd_data_flow1

It is actually not too terribly bad. When you add the Slowly Changing Dimension a wizard pops up and when all the stars align, all the data flow transformations and destination below are created.

If we focus on the top of the data flow first, it is easy to see I am pulling from two source systems and doing a union all. The interesting problem I had to solve was the deleted VM problem. The wizard didn’t do that for me. I knew if I had the staging table, I could compare that to the dimVM to see if anything was missing. If you want to find out what is missing, use an EXCEPT query. Once you find out what is missing (deleted VMs) we can update the validto field effectively closing up shop on that row but keeping the history of rows relating to that VM. I decided to add the isdeleted column to make it easier to find deleted VMs. This code is in the SQL Script task on the control flow.

update dimVM
set dimVM.validto = getdate(), dimVM.isdeleted = 1
from dimVM
inner join (
select vmid,vcenter from dimVM
where validto is null
except
select vmid,vcenter from dimVM_staging
) del
on dimVM.vmid = del.vmid and dimVM.vcenter = del.vcenter

One last little tidbit. If you make any modifications to the transformations that the SCD wizard created, you should document them with an annotation. If for some reason you have to get back into the wizard, it will recreate those transformations from scratch… ironically not maintaining any history.

Step 3. Profit

I hope you enjoyed hearing about my new experiences in the Slowly Changing Dimension transformation in SSDT.

 
1 Comment

Posted by on April 14, 2015 in SQL Dev

 

Final Preparation for 70-463 Implementing a Data Warehouse with Microsoft SQL Server 2012

This is a continuation of this post

Two fellow bloggers have been posting more quality information on this test.
http://colleenmorrow.com/

http://www.desertislesql.com/wordpress1/?p=243

When reading the book I skipped over all of the practice sections. I did read the exam tip sections inside of the practice but never actually practiced. I don’t have a lot of hands on experience with SSIS and even less with mds/dqs. I spent about 9 weeks making through the book while skipping the practice and most of the reviews. I probably would have needed an additional 18 weeks to properly make it through all of the practice or lab type sections of the book. Learn one, do one, teach one is my favorite method to mastery but with 2nd shot deadline, I didn’t have a lot of time to prepare.

To supplement, I attempted to find videos on youtube and watched videos on the Microsoft Virtual academy. Both sources were not very demo heavy. What I did find is CBT nuggets that give a 7 day trial. The 70-461 videos that I was able to watch were very high quality, fast paced and demo heavy. This is exactly what I needed at this time. I’d recommend a membership if you have a bundle of money burning in your pocket.

Since my trial was up I decided to type up my CBT nugget notes.

CBT connections managers
control flow -> doesn’t involve data
bottom level are private connection managers, a.k.a package level
right solution explorer is project level connection managers which are global
you can enable/disable sequence containers
precedence constraints, go to properties to define AND or OR logic
copy-> paste package connection managers
delay validation -> doesn’t check structure
email doesn’t have a port option but could purchase add-ins or write your own
fix for NULLs is COALESCE

Data Flow
rows, buffers, pipeline,transformations
raw file -> ssis only -> good for sharing data between packages
raw file -> good for resuming packages
recordset->variable used to loop through
for performance, aggregate at the source since that is blocking
import export col -> for blob data
term matching is like CTRL+F
blocking tasks take lots of memory -> sort, aggregate
partial-blocking -> merge chuncks

Data Quality services
cleansing matching
server is 3 databases
dqs client is used for creating KBs
creating a knowledge base
-open xls sheet -> job title list for KB example
-KB needs a domain, circle with * button is domain
State length of 2 is an example domain rule
composite domain (EX: address which includes city state zip)
reference data source RDS (ex: mellisa data for addresses)
KB’s get published
activity is automatically logged

Implementing DQS
data profiling task in SSDT
-profile types
–null ratio request
–pattern generator RegEx for formatting
–column statistics
-then specify column
Quick profile: runs against all columns
Open data profile viewer
Cleansing
suggested confidence level
corrected confidence level
DQS cleansing task
Job title source job_title _output
jobtitles table
matching
newKB->domain->source column (survivor record)
the table with the + button to add a rule and use the Rule Editor

Implementing MDS
proactive management
people place concepts or things
non-transaction data is good for MDS
includes auditing and versioning
MDS Componenents(Database, config mgr, MD mgr, web service, mds model deploy, excel Add-In)
MDS Objects(Models: the container db, Entities: like tables, Attributes: like columns, Hierarchies, Members: Actual data)
Install requires powershell 2.0 and IIS 7.5, silverlight and a database
has integration with DQS
to deploy packages that contain data must use CLI (deploynew -package “” -model)

Data flow
merge join requires sort -> advanced editor, pick isSorted and the column
MetaData problems: double click on flow and change types
Lookup transformation
-cache connmgrs for re-use
–redirect rows
–multi output popup
slowly changing dimension task (wizard)
fixed attribute fail on change
changing attribute type 1 overwrite type 2 new records (history)
inferred member flag goes in dimension
blocking oledb command
redirect error rows to flat file

executing packages
“$Project::ProjectParam(String)”; test@email.com
LOGGING_LEVEL 3 = Verbose
dtexec.exe is fire and forget style
xp_cmdshell
built-in SPs in ssisdb
catalog.set_obj_param value
restartable packages
-checkoint file
-tracking last successful step in control flow
project properties
-select file name
-set usage never
–if exist
-save checkpoints = true
-set property fail package on failure = true
to test, can set task property to force a failure

 
Leave a comment

Posted by on April 9, 2015 in SQL Admin, SQL Dev

 

More Preparation for 70-463 Implementing a Data Warehouse with Microsoft SQL Server 2012

This is a continuation of my previous post. This is just some very quick notes that I am posting for my benefit and so that readers may get an idea of the preparation necessary for this test. They are my notes from this book: https://www.microsoftpressstore.com/store/training-kit-exam-70-463-implementing-a-data-warehouse-9780735666092

PART II: Developing SSIS Packages

simple data movement – can use import export wizard
complex data movement – SSDT
SSDT is visual studio shell used to develop IS,AS,RS projects
Control Flow connection managers can be package or project scoped
Connection manager types:
ADO – backwards compatibility
ADO.net – compatible with sql server
AS – analysis services
File – SSIS data type
Flat file – delimited file
ftp – security option is only basic auth
http – web services or file, no windows auth
OLE DB – sql server, will be removed in favor of ODBC
ODBC – open database connection
SMTP – basic email auth only

package scoped connection managers will override the higher level project scoped connmgrs

control flow tasks and containers
containers help control execution of tasks
transformations include
cleansing – remove invalid data or unwanted data
normalization – XML value to varchar
conversion – byte[] to varbinary(max)
translation – “F” to “Female”
validation
data calculation and data aggregation
data pivoting and data unpivoting

ssis tasks categories, data prep, workflow, data movement, SQL admin, SQL maintenance

containers, for loop, foreach loop, sequence

Precedence Contstraints ( the arrows that come off of tasks)

success, failure, completion
dotted lines mean OR and solid means AND logic used when multiple tasks are involved in flow

Designing and Implementing Data Flow

Data Flow is a level deeper than the control flow
Control flow triggers data flow
data flow task builds execution plan from data flow definition
data flow engine executes the plan
*Validate external metadata – checks for existance of tables and objects and should be turned off if they are dynamically created
builk OLEDB = fast load
ODBC = batch
fast parse is available at the column level on some data types ( date, time, int )
Working with data flow transformations
-Blocking (ex: sort, aggregate) transformations that read all data in before passing any rows down the pipeline
-Non-Blocking -> lookup, multicast, conditional split or other row-by-row transformations
-partial-blocking -> merge, merge join, union all, data flows in chunks
cache transformations – good for multiple transformations on same data
import/export col – good for blobs
character map – upper case, lower, linguistic bit operations
advanced data prep: dqs cleansing, oledb command, slowly changing dimension, fuzzy grouping, fuzzy lookup, script component(custom.net)
#NEW# Resolve references editor helps resolve mapping problems
Lesson 3: strategy and tools
lookup transformation caching
how to handle rows w/ no matches
sort is expensive, optionally perform sorts at source and use advanced editor to mark data as sorted
avoid update and delete on fact tables
do large table joins on database layer
do updates on loading or temp tables in set based sql operations
Chapter 6: Enhancing Control Flow
ssis variables and parameters
avoid retrieving external source variables more than once
parameters are exposed to the caller bu variables are not
parameters are read-only and can only be set by the caller
variables are helpful to reuseability
variables are user defined or system
variables can store rows foreach enum containers
-avoid storing large rowsets in memory/variables
variable data types
-object: last resort
Int16: -32,768 thru 32,768
UInt16: 0 thru 65,535
UInt32: 0 thru 4,294,967,295
DateTime
String
Char: 65,353 unicode
Decimal: 28 or 29 significant digits
Single
Double
Variable Scope
-Package Scopre
—-Container Scoped
——–task scoped
property parameterization
explicit assignment
expressions
lesson 2: connmgr, tasks, and precedence constraint expressions
expression: combination of constants, variables, parameters, column refs, functions, and expression operators
-special ssis syntax close to C++
math functions: ABS, EXP, CEILING, etc…
String Functions: FINDSTRING, HEX, LEN, LEFT, REPLACE
precedence constraints can use AND/OR logic expressions
Lesson 3: Master Pakcage
just a normal package that uses the execute package task
use variables to expose results to parent
use project deployment model to make parameters available to child packages
use project scoped parameters
CHAP7: Enhancing Data Flow
Lesson 1: Slowly Changing Dimesions
-late arriving dims or early arriving facts
–1. insert row into dim, mark inferred… requires bit col
–2. use newly created surrogate key
–3. when loading dim overwrite inferred members
TYPE 1 SCD: overwrite
TYPE 2 SCD: keep all history
can use conditional split to see what columns changed
ex: source.fullname dest.fullname
using t-sql hashbytes can compare for changes
–then two cols for hash val Type1 & type2
use set based updates instead of wizard
Lesson 2: preparing a package for incremental load
dynamic sql
change data capture
Dynamic SQL in OLEDB source
1. select dataaccess mode of sql command and use ? to pass parameter
2. pass variable to sql command and use expressions to modify the sql string
cdc functionality – cdc source and cdc splitter
-ALL, ALL w/old, net, netw/update mask, net w/merge
lesson3: error flows
route bad rows – fail, ignore (copies null), redirect rows
chapter 8: creating robust and restartable packages
can set transactions at package control flow or task level
transactions use msdtc
transaction options are: required, supported, not supported
transactions work on control flow not data flow
can nest a not supported execsql that won’t rollback inside a transaction (ex: still want to audit on fail)
lesson2: checkpoints
save checkpoints need turned on, on package
creates a file and restarts if exists
starts from begining if not exists
lesson3: event handlers
onerror
onwarning
onvariablechanged
ontaskfailed
can turn event handlers off for task
chapter 9: implementing dynamic packages
project level and package level connection mgrs and paramters
must be deployed to ssis catalog
parameter design values are stored in the project file
cannot change parameter value while package is running
property expressions are evaluated on access
lesson2: package configs
enable package deployment model
can get parent package configs
chapter10: auditing and logging
logging: package configuration
auditing: dataflow trnasformation component
lesson1: logging packages
providers are: txt file, sql profileer, sql server, event log, xml
boundry progress exception
use parent setting is default
ssis control flows can be configured for logging
lesson2: auditing and lineage
elementary auditing – captures changes
complete – adds usage or read activity
audit transformation editor
lesson3: preparing package templates
keep packages in source control

Part IV: managing and maintaing ssis packages
ssis service is required in production
ssisdb new
package install utility is legacy
can use ssdt or ssms to deploy packages
project model or package model
dtexecui is legacy
can use TSQL, powershell, manual dtexec cli to execute packages
agent to schedule packages
introduced master package concept
securing packages: uses sql security concepts of principals and securables
ssis_admin role
ssis_user by default allowed to deploy, and deployer is allowed to read, modify, execute
Chapter 13: troubleshooting and perf tuning
breakpoints work only in control flow
breakpoints and fire on a hit count
data viewers on path will show grid view of data
use error outputs to catch bad rows
test with a subset of data
basic logging is default
switch to verbose when there are problems
data taps are like dataviewers for production
must be predefined using catalog.add_data_tap for specific data flow
lesson2: perf tuning
buffers are a group of data in data flow
determined automatically
Transformation Types
-non-blocking: row based synchronous
-partial blocking: asynchronous transformation
-blocking: asynchronous
backpressure controls flow for best memory control
max buffer rows – 10,000 default
max buffer size – 10MB by default
fast load on destination
full-cache lookups
avoid oledb transformations
BLOBs get swapped to disk
data flow engine threds
max concurrent executables -1 = # of logical processors +2
perfmon counter: buffers spooled

PART V: Building Data Quality Solutions

chapter14: installing and maintaining DQS
Soft dimensions: timeliness, ease of use, intension, trust, presentation quality
hard dimensions: accuracy, consistancy
Schema dimensions: completeness, correctness, documentation, compliance w/theoretical models, minimalization
activites: understand sources and destinations
lifecycle
security and backups managed through ssms
Chapter15: implementing MDS
metadata, transactional, hierachical, semi-structured, unstructured, master
MDM goals: unifying or harmonizing, maximize ROI through reuse, support compliance, improving quality
MDM: coordinated set of tools policies to maintain accurate master data
map master data dimensions to DW
Installing MDS: DB, Service(Needs IIS), Manager, Excel Add-IN
Creating MDS model
1.Model
2.Entities(like tables)
3.Attributes(like columns)
4.Hierarchies
5.collections
Derived hierarchies: Recursive with TOP = NULL (ex: Org Chart)
Explicit Hierarchies – Organization can go any way
Collection: flat list of members
MDS service performs business logic
Chapter16: managing master data
MDS Packages
-Model deployment package to move data to another server
-wizard only includes meta data
-permissions are not included
-MDSModelDeploy command prompt if you want to move data
exporting – tsql on subscription views, webservice
Security, system admin (one user, tsql to change), model admin (complete model access)
entity permissions apply to all attributes
mds add-in for excel (connect to to http://server:8080)
when model and member permissions are overlapping read-only > updated and deny > *
excel add-in can use DQS KB matching
Chapter17: creating a data quality project to clean data
knowledge disovery
domain managment
reference data services
matching policy
domain: semantic representation of column
properties: data type, leading values, normalize, format, spellchecking
Term basic relation: Inc. -> Incorporated

I skipped 18,19,20: Advanced ssis and data quality topics because only 5 parts are listed on the exam prep and I ran low on time.

 
1 Comment

Posted by on April 8, 2015 in SQL Admin, SQL Dev

 

Preparation for 70-463 Implementing a Data Warehouse with Microsoft SQL Server 2012

I’m writing this post to force myself to spend some quality time with the materials for this exam. I have been at it for almost two months now and am nearing my exam date. I accelerated my plan so I could get into the 2nd shot window offered by Microsoft and also so I could finish my MCSA within 1 year. It has been a battle at times and is not an easy certification to get. Microsoft has really increased the difficulty since the MCITP for SQL 2008 which only required 2 exams.

My employer is assisting with the costs in a few ways. They will reimburse me for the cost of a passed exam. They are giving me a $500 bonus if when I pass all three exams and prove my MCSA. And they have loaned me the Training Kit book along with the other tests books that I have already returned.

My plan has been going fairly well. I’ve been able to put at least 10-15 minutes in about 6 days a week. Some of those have lasted and hour or more but that is pretty rare. Data warehousing is interesting to me because we have a lot of things starting up at work that may take off and require these skills. Before I started studying I had deployed only a few packages for my own small data collection and reporting tasks as an administrator. I also do not get too involved with database design since we rely on a lot of 3rd party applications. That world is changing for me and that is why I have been able to be a fairly good student for this last test.

So lets get to my plan.

The percentages are the first thing to note on this page: https://www.microsoft.com/learning/en-us/exam-70-463.aspx

11% – Design and implement

23% – Extract and Transform

27% – Load

24% – Configure and deploy SSIS

15% – DQS

_______

100%

I like to sit down with the book and read as much as I can while taking notes. I write down a lot. When I look at it later I think, “duh I knew that why did I write it down?” But it actually helps me stay focused. Even if I just write down the title of the section, it keeps me on track. At this point, I am ready to go back and review a lot of those notes and type them up so here they are.

The book is split out into those same 5 “Parts” as listed on the exam website.

Part 1: Design and Implement
Use snowflake in a POC since it will be easier to design from the complex OLTP environment.
Star schema for everything else.
Star is just a simplified, denormalized, merged, cleansed, historical schema with fewer joins
Star schema works well for SSAS cubes, SSAS won’t be on the test (phew).
A fact is: “Cust A purchased product B on date C in quantity D for amount e”
Dimension table: Customer, Product, Date
One star per business area
The Granularity level is the number of dimensions or depth you can slice by (thinks sales by quarter or sales by day)
Auditing: Who, What, When
Lineage: Where is the data coming from?
Dimensions: The goal is to make it look good in a pivot chart
-descretizing: putting values into bins and not keeping too much granularity because it doesn’t graph well
-Member Properties: columns not used for pivoting
Slowly changing: type 1- no history, overwrite; type 2 – keep history with current flag or validto-validfrom cols; type3 – limited history with additional cols like prevAddr
Keep business keys intact, create additional DW specific keys (surrogate keys) for linking fact to dimensions, probably INDENTITY
Use a SEQUENCE if you need to know the number before inserting, request multiple at once, or need a multi-table key
FACT TABLES: made up of FKs, Measures, Lineage cols, Business keys
consider the additivity of measures. EG: can’t sum an AvgDiscCol
Fact tables should be on the Many side of the 1->many relationship
Dimensions contain the lineage data
Age is a common computed column
design dimensions first, then fact tables
use partitioning on your fact table
Fact tables contain measures
Every table should have a clustered index
Do not index FKs of fact table because HASH joins dont need it?
If you are doing merge joins and nested loop joins indexes on FKs help
indexed views are useful in some cases
Row/page compression automatically applies unicode compression
batch mode is faster and will show in the query plan
column store indexes: one per table, not filtered, not on indexed views
Partitioning function maps rows to a partition
partitioning scheme maps partition to filegroups
aligned index: table with same schema which allows for partition switching
optimizer can eliminate partitions
inferred member: row added in dimension during fact table load

PART II: Developing SSIS Packages
To be continued…

 
2 Comments

Posted by on April 2, 2015 in Uncategorized