Innovation. Imagination. Delivered. RSS 2.0
# Tuesday, April 22, 2008

I am out at DevConnections this week and so far I have been impressed with the improvements and new features that are coming in the 2008 product line. I will post recaps of the sessions I attended as time allows...

- Data Warehousing in SQL Server 2008 by Christian Kleinerman – this session focused primarily on the new features and performance enhancements that were made in the SQL engine.

ETL Enhancements:

1. The new MERGE statement – this powerful new statement combines DML statements into one to allow for “UPserts” and “Deletes” between a specified source and destination.

MERGE tblDestination t

USING tblSources s

ON t.ID=s.ID

WHEN MATCHED

UPDATE SET t.Name=s.Name

WHEN NOT MATCHED

INSERT VALUES(s.Name)

WHEN NOT MATCHED BY SOURCE THEN

DELETE

2. Logging enhancements were made to allow for minimal logging. This level of logging keeps only what is needed to successfully rollback. According to there test significant performance gains where seen with this change.

3. Insert over DML – you know have the ability to do an insert off a DML statement with the OUTPUT clause.

INSERT INTO tbl

SELECT FROM (

<>

) CHANGES

WHERE Action=’UPDATE’

4. New tabled-valued functions were attended to track and query changed data.

5. The data flow thread allocation was changed from static to dynamic assignment for operators. This improves parallelism performance and adds thread starvation prevention.

6. A new ETL lookup is being added to SSIS. The new lookup has an unrestricted cache size and the ability to persist cache.

7. Data Compression to shrink table sizes and improve query performance is being added. The compression can be configured at the table or index level and they have seen performance gains anywhere between 1.5x and 7x improvements. During this topic they Mr. Kleinerman discussed the tradeoff between memory and CPU usage when using compression. He also briefly covered how compression was implemented (I will go through later).

8. Enhancements were made to the designer to build ‘best practices’ into the tools suite.

Query and Analysis:

1. Partitioned Table Parallelism was improved through dynamic thread allocation.

2. SQL Server 2008 can support multiple bitmap filters to improve star join query processing.

3. GROUPING SETS where added to the sql toolbox

GROUP BY GROUPING SET (date, (state, city))

4. Changes were made to the MDX query processing for subspace computation. Since cubes are sparsely populated changes were made to minimize what is processed during computation.

5. The Reporting Services engine was completely rewritten for 2008. I won’t go into all the changes but there were a ton of them discussed (i.e. Enhanced visualizations, Word Rendering, etc).

Admin

1. Backups can not be run with compression

2. The entire AS backup storage system has been replaced

3. The Resource Governor basically allows workload to be differentiated and per-request limits to be set ( I plan on elaborating on this is a separate post).

Tuesday, April 22, 2008 8:36:28 AM (Eastern Daylight Time, UTC-04:00)  #    Comments [0] - Trackback
Conferences
# Tuesday, March 11, 2008
I had an opportunity to attend my first Tampa SQL BI Meeting last night and I wanted to share some of my thoughts from the SSIS demo that was given.
The first thing that really caught my attention is complexity of expressions that are used in a conditional split. I started picturing the nightmare scenario of debugging it and figured there has to be a better way. Enter the Script Component. The script component allows you to define multiple outputs and by monkeying (yeah thats a technical term) with the ExclusionGroup property you can filter the input rows. I am probably biased towards this technique (since my background is primarily in software development) and will need to find the time to do some analysis to see if there is a performance difference between the two methods.
The second item that got my thinking during the demo is more of a difference in approach. The case that was being presented had a requirement that before the package did any work a rowcount on the source data had to occur. If the row count returned 0 no further processing occured otherwise the destination was truncated and reloaded. To accomplish this task the presenter used an 'Execute SQL Task' to run the count query and then used to a 'Script Task' to determine if the control flow should proceed.
My thought on accomplishing this 'requirement' is a little more condensed. I would use the 'Execute SQL Task' but instead of a 'Script Task', I would use a 'Precedence Constraint' to modify the flow if necessary. By writing a simple expression to determine if the variable assigned to in the 'Script Task' is greater than zero you could either continue with required work or direct to an error flow.
Thats pretty much all for today...


Tight Lines!!


--Chris
Tuesday, March 11, 2008 7:33:53 AM (Eastern Standard Time, UTC-05:00)  #    Comments [0] - Trackback
SSIS
Archive
<April 2008>
SunMonTueWedThuFriSat
303112345
6789101112
13141516171819
20212223242526
27282930123
45678910
About the author/Disclaimer

Disclaimer
The opinions expressed herein are my own personal opinions and do not represent my employer's view in any way.

© Copyright 2012
Christopher L Price
Sign In
Statistics
Total Posts: 13
This Year: 0
This Month: 0
This Week: 0
Comments: 0
Themes
Pick a theme:
All Content © 2012, Christopher L Price
DasBlog theme 'Business' created by Christoph De Baene (delarou)