BI/BPM - The SeeQueL

Lists of Database and BI Tools

2019-08-19T05:30:00.001-07:00

Some utilities for SQL, let's hope you never need to use the Microsoft Kerberos Configuration Manager.

Free SQL Tools
https://www.sqlshack.com/free-sql-tools/

Searching Github for awesome always brings up some great lists.

Awesome SQL Tools
https://github.com/danhuss/awesome-sql

Awesome BI Tools
https://github.com/thenaturalist/awesome-business-intelligence

Awesome Databases
https://github.com/agarcialeon/awesome-database

And the greatest is.... Konstantin Ktaranov's SQL Server Kit
https://github.com/ktaranov/sqlserver-kit

Windows 10 Update Black Screen of No Return

2019-06-13T03:18:00.000-07:00

I managed to recover from a black screen after login issue with the Windows 10 April 2018 update. Here's some of the troubleshooting steps taken to apply a later update to fix it.

1. unplug the 3 Saitek AV8R joysticks I bought last weekend from a yard sale for $8. Didn't help!
2. login to a black screen with cursor.
3. ctrl-alt-del and open task manager
4. start - run - control panel
5. uninstall AVG Antivirus
6. start - run - chrome
7. search for Windows 10 upgrade
8. download Windows10Upgrade9252.exe from Microsoft
9. install Windows Build version 1903 18362.175.

https://www.windowscentral.com/how-download-and-install-windows-10-cumulative-updates-manually
https://www.drivethelife.com/windows-10/windows-10-black-screen-after-creators-update.html

After a couple hours the machine is logging in. Still unclear about stability. List Windows Update shows I have no updates available. System Restore was turned off, so turned it back on and created a restore point.

Diagnosing:

From Task Manager, I ran eventvwr to get the latest events. There was an error with AVG and and fault in Explorer.exe after the previous update which seemed to indicate the virus scanner may have blocked the update. Start - Run - Explorer usually brings back the start menu - this time it was faulting on LiveTile.dll. Sfc /scannow from an administrator prompt did absolutely nothing.

There are still some issues. Checking eventvwr and selecting filter-events crashes the snapin.

The process cannot access the file 'C:\ProgramData\Microsoft\Event Viewer\Views\View_0.xml' because it is being used by another process.

The link below helps to fix this. In my case I just renamed the 2 views_*.xml files to views_*.xml.bk

https://www.ghacks.net/2019/06/12/windows-10-event-viewer-error-after-installing-kb4503293-and-kb4503327/

At least I'm only a few days behind everyone else now.

Fixing OneNote 2013 Mouse Clicks

2018-05-14T19:30:00.000-07:00

Microsoft tools are often intertwined with the Tools - Internet Options settings in Internet Explorer. Oddly enough, a Windows 10 or OneNote update caused OneNote 2013 to prevent mouse clicks inside a note. Right-clicking wouldn't work either.

The fix? Tools - Internet Options - Connections - LAN Settings - Turn off Automatically Configure Settings, Turn on Use Automatic Configuration Script, hit ok.

Worked immediately for me.

Cron Scheduling and Nano Services in Open Source Azure Functions

2016-10-21T04:23:00.002-07:00

There are many ways to schedule jobs in Azure. An interesting and fairly simple one available is to use Azure Functions, a lil' brudder of WebJobs. For up to 1 million 1 second functions, there is no cost. You can schedule a job function in Bash, C#, F#, NodeJs, Python, Java, etc.
https://azure.microsoft.com/en-us/documentation/articles/functions-bindings-timer/

Example json for a function. http://www.crontab-generator.org/ for those who have not mastered the art of reading job schedules in Cron, this will run at 3:25am UTC every Monday - Friday.

In this case, your "crontab" or cron table part of the JSON formatted job definition.

{
  "bindings": [
    {
      "schedule": "25 3 * * 1-5",
      "name": "nightlyReports",
      "type": "timerTrigger",
      "direction": "in"
    }
  ],
  "disabled": false
}

This is the function example to log the timer event.

public static void Run(TimerInfo myTimer, TraceWriter log)
{
    log.Info($"C# Timer trigger function executed at: {DateTime.Now}");    
}

If your job is to execute a SQL stored procedure such as an SSIS proc to kick off an ETL package, or archive / groom data.
https://azure.microsoft.com/en-us/documentation/articles/functions-scenario-database-table-cleanup/

Another example is an event-based trigger coming from an Event Hub, or Blob Storage based trigger coming from a blob, table, queue, or file drop.

Once a function has been created, you can test or run the function with cURL by copying the Function Url.
https://azure.microsoft.com/en-us/documentation/articles/functions-test-a-function/

The code is in GitHub, packaged as the Azure Webjobs SDK Script.
https://github.com/azure/azure-webjobs-sdk-script

Announced at Build 2016, Some facts on Azure Functions from MSFT Playground
https://msftplayground.com/2016/05/facts-azure-functions/

Do you use Azure Web Jobs or the new Functions Preview?
https://blog.kloud.com.au/2016/09/14/azure-functions-or-webjobs/

Functions Reference
https://azure.microsoft.com/en-us/documentation/articles/functions-reference/

Functions with Logic Apps
https://azure.microsoft.com/en-us/documentation/articles/app-service-logic-azure-functions/

Continuous Deployment
https://azure.microsoft.com/en-us/documentation/articles/functions-continuous-deployment/

Machine Learning with R and SQL 2016 / RevoR + Open Source MSFT

2016-10-19T17:50:00.000-07:00

Ginger Grant, Data Platform MVP discusses Machine Learning with R and some Custom Reports for SQL 2016 R Services.

Everything that exists is an object.
Everything that happens is a function call.
-- John Chambers, Advanced R

Microsoft Open R is the enhanced distribution of R, which provides the ability to create your own enterprise MiniCRAN, reproduce results between package version snapshots, parallelism/multithreading, and is portable across Windows, Linux, and OSX. Enterprise is constantly looking for stability, and https://mran.microsoft.com/faq/#change-repos provides this with fixed snapshots across all OpenR users.

RStudio hopped on the Jupyter Notebook bandwagon a couple months ago.
http://rmarkdown.rstudio.com/r_notebooks.html

In 2017, we will have SQL Server 2016 with Microsoft OpenR and R Notebooks, running on Red Hat Enterprise Linux along with Hortonworks HDP...

Getting Started with PHP 7 SQL Server and Azure SQL Database on Linix Ubuntu with Apache

And something completely unrelated, perhaps useful, Yet Another Network Tester, with multithreading and CPU affinity.
https://github.com/Microsoft/ntttcp-for-linux
https://gallery.technet.microsoft.com/NTttcp-Version-528-Now-f8b12769

Every Microsoft Github on a single page

2016-10-18T23:02:00.000-07:00

Microsoft's foray into the land of *nix is 20 years in the making

2016-04-13T20:33:00.001-07:00

Announced at the Build 2016 Conference in March. We are back in the days of one of my specialties, command line OSes and command scripts. Bash comes to Windows. Ubuntu runs on Windows.

This isn't really that new, is it? Utilities and SDK for Subsystem for UNIX-Based Applications (USDKSUNIX-B?) is the unfortunate name of GNU utilities on Windows, formerly known as SFU 1.0, released in 1999. Subsystem for UNIX-based Applications? So much for catchy names like Tinkerpop, Flume, Kafka, Storm, Solr, Spark, Hive, Pig, YARN, or Mahout. And oh yes, Java. Hasn't Java had portability across platforms for awhile?

The efficiency and performance of Ubuntu on Windows is really what is impressive here.

Why is this happening? Did Microsoft give up on Powershell? Is Powershell coming to Linux? Is Windows 11 going to run on SPARCstations? I am so confused...

For those making the transition, a helpful MS-DOS - Bash help sheet, courtesy of Red Hat.
awk, ed, sed, grep, top, locate, tail, head, cat, vim, perl, ssh...

Apparently 60% of Azure is now running Linux. At least 15% of the world's power is being consumed by data centers. A UNIVAC consumed around 67 kilowatts. The world consumed around 105,000 Terawatts in 2012. Power is a huge driver in terms of cost and efficiency is super important when a server is running 8,760 hours / 365 days a year. Using a minimalist operating system (or none at all?) could help with additional savings.

It is the year two thousand and sixteen, the 21st century. Shouldn't we be celebrating our separation from monitors, mice and keyboards with augmented, virtual reality, neural impulse actuators, heads-up displays and immersive operating systems that feed us information via telepathy?

Apparently there is some delay, though it's not caused by Microsoft. Hololens is shipping now. Oculus, not so much.

I bought an OCZ NIA brain mouse a few years ago. It has been collecting dust ever since I decided I had better things to do than to train my brain to play pong. My eyes hurt from blinking to shoot in Unreal Tournament, turning my face away from the screen to look around was a bit counter intuitive, and the grounding strap required to keep signals level didn't add to the experience. Plus I still needed my mouse and keyboard.

It is still impressive to me that we have the technology today (and 8 years ago!) to read brain waves for under $100. And we now have the technology to write them using transcranial direct current stimulation (tDCS), for the low low price of $39.95!

And to think that Nikola Tesla used to squish his toes together 100 times a night to stimulate his brain cells, when a 9-volt battery and a wet sponge could have done the same trick.

So what's next? SQL 2016 on Linux?
http://blogs.microsoft.com/blog/2016/03/07/announcing-sql-server-on-linux/

Free SQL Licenses for Oracle customers?
https://www.microsoft.com/en-ca/server-cloud/sql-license-migration.aspx

In 1998, Cygwin 2.0 introduced the Windows world to GNU tools.
http://www.math.utah.edu/~beebe/gnu-on-windows.html

In 2016, oh my zsh comes to Windows...
https://twitter.com/ohmyzsh

Tomorrow, I'm installing this!
sudo yum install byobu -y --enablerepo=epel-testing

Convergence was the buzzword a few years ago. This year's buzzword should be... convergence?

Well, back to reading copies of Computerworld magazine from 1997 for new product ideas... did you hear PC market leader Compaq is pushing their new Windows NT models as a Unix alternative?

Microsoft goes All-In on Analytics

2015-10-29T22:46:00.000-07:00

It has been a busy few months in the data analytics space, especially at Microsoft. To say the least. Next year will be even busier, when many of these solutions converge further.

Microsoft announced General Availability of PowerBI, and brings many of its cloud features to the desktop.

Jen Underwood reveals the splendor in PowerBI Desktop. PowerBI Desktop can be published to a Pyramid Analytics server for on-prem solutions. PowerBI isn't just in the cloud anymore.

Microsoft acquires Datazen, and some multidimensional dashboards come out of a DataZen server in SQL 2016.

SQL 2016 with R & JSON integration. Hold up, Revolution Analytics is coming to SQL? JSON in SQL? No more FOR XML madness?

New JSON support in T-SQL, the SQL Server query language, includes OPENJSON to parse JSON text and return tabular data, JSON_VALUE and JSON_QUERY for querying JSON data, and ISJSON for validating JSON-formatted text.

The syntax for the R external functions seems suspiciously familiar to what Apache Hive provides..

PIP INSTALL JUPYTER

ML Blog Team talks about Jupyter Notebooks for Azure ML Studio and Nitin Mehotra gets us started with Zeppelin Notebooks and Apache Spark on Azure HDInsight.

Michael Erasmus gives us an interesting mashup of Python and SQL running in a Jupyter notebook.
http://michaelerasm.us/redshift-and-ipython-bffs/

What's nice about this is the inline SQL statement, effectively removing the need for quoting & appending multi-line SQL together. What else is useful is the ability to capture the results of the SQL query as a data frame for analysis and visualization.

A similar method might be used with MDX queries and Analysis Services OLAP cubes.
http://bkanuka.com/articles/python-connect-to-olap/

Perhaps if you have some web services connected to Excel Services or PowerPivot, this method could get really useful for exposing your data. I would imagine writing some DAX queries nicely formatted in Jupyter python and returning data frames to R would be of interest for a couple people... well maybe just me. ;)

Speaking of DAX...

Kasper de Jonge writes about SQL 2016 includes over 50 new DAX functions
http://blogs.msdn.com/b/analysisservices/archive/2015/09/02/what-s-new-in-microsoft-sql-server-analysis-services-tabular-models-in-sql-server-2016-ctp-2-3.aspx

Some of the cool ones are Natural Inner Join,

CONCATENATEX(Employees, [FirstName] & “ “ & [LastName], “,”)
Returns "Alan Brewer, Michael Blythe"

SUBSTITUTEWITHINDEX (Left Semi Join in PowerPivot?)

ADDMISSINGITEMS (What do you mean we have no sales for March in Tuvalu? We need to see a March column for Tuvalu on the report!)

Only 46 functions left to check out...

Unit Testing and Mocking T-SQL Objects with tSQLt

2015-10-13T18:33:00.000-07:00

tSQLt is an open-source framework for unit testing SQL Server code. You can get it here. The Red Gate SQL Test application provides an installation, management framework, support, and testing harness for SQL Server Management Studio. ApexSQL has also announced a competing offering, to be released in 2016.

The most useful functionality, in my opinion, is the ability to mock or fake objects. See Greg Lucas' post "A mock too far?" for some useful scenarios. There are many other tSQLt tutorials and examples on the DataCentricity web site for Test Driven Database Development (TD3) using mocks.

What do you mean, mock or fake objects?
When running a test or test suite with a SQL testing framework like tSQLt, you arrange your test, act on the item under test, and assert your results match expectations. The framework handles the assertions, teardown and rollback of any items changed by the test. What mocking allows for is the faking of dependent items not central to the purpose of the test. It provides a shell or template object that is a clone of an existing object without the clutter of reams of data, foreign keys or constraints unrelated to the test.

tSQLt allows you to fake tables. Behind the scenes, tables are renamed, constraints are removed and all columns are set to nullable. This allows for you to assert / check only the columns and rows of data required for your test. Instead of sifting through millions of rows looking for your changes, you can assert that your input row was added to the destination table with a simple count. You can create an expected table, and use tSQLt's AssertEqualsTable to compare that the columns you are testing contain your expected results. You can take your expected results from a table, fake the table, then ensure that the results are repeated after execution of the object under test.

It's kind of like Where's Waldo, with only a white page and Waldo sitting in the middle of it. It becomes very obvious when things are and aren't working, since the noise has been removed from your test.

tSQLt allows you to spy on procedures. Behind the scenes, procedures are emptied out, and only the shell or "contract" remains. The parameters of the procedure remain, with an insert into a logging table. This allows you to inspect what is being sent to procedures to ensure the item under test is passing along the expected parameters. It also allows you to inject code into dependent procedures, perhaps to test functionality or changes that would exist in the future once that other team gets finished with their deliverable.

These types of actions follow the Test Double pattern.

How can I use this when debugging SQL code?
Fake out and spy on all dependent objects to get a better understanding of how the procedure under test is acting against them.

I have found Faking and Spying are also useful when debugging and understanding poorly documented procedures, since you can interactively view your data during a test, roll back the changes, and visualize what occurred in the destination table. If you are working with a large, legacy procedure or set of procedures, this tool becomes invaluable for tracing data lineage and transformation logic. A side effect is you'll probably end up with some automated unit tests on your legacy code.

Martin Fowler describes these mocks or fakes as a kind of Stunt Double, taking the bombardment of your tests while your true actor is sitting on the sidelines.

In the past, I used synonyms for similar purposes, however creating fakes for testing rather than replacing synonyms is much a much easier task with tSQLt.

If you are using SQL Server for more than just a big dumb ORM storage or CRUD layer, it's worth looking at a unit testing framework such as tSQLt and the ecosystem around it.

Seb Rose has a great article around Behavior Driven Database Development (BD3) including examples of Test Data Builder and Data Mother patterns for creating test data.
Ed Elliott's SSDT & tSQLt - a Serious Double Act and Generating tSQLt Tests from a DACPAC shows how Visual Studio SQL Tests can incorporate the tSQLt framework, and how you can generate test stubs from procedures in a DACPAC.
The tSQLt Google Group for questions on the framework
Denis Gobo gets you setup with tSQLt and walks through the basics.
Steve Jones has some good Youtube videos on the topic, and a recent article about hacking your way into tSQLt on Azure.

Reporting Services 2016

2015-09-13T19:02:00.001-07:00

After a couple years hiatus, the SSRS Team (via Riccardo Muti) has published some highlights of SQL Server Reporting Services 2016 CTP 2.3

http://blogs.msdn.com/b/sqlrsteamblog/archive/2015/09/02/what-s-new-in-reporting-services-in-sql-server-2016-ctp-2-3.aspx

Most of the highlights are for usability and appearance. No Metro tiles?
http://sqlkover.com/report-builder-changes-in-sql-server-2016/

Rather than SSRS features, my bigger interests within SQL 2016 Enterprise are Polybase and the SQL-R functionality. Polybase will allow you to join your Data Lake with the Data Ocean that is your Enterprise Data Warehouse, and all things in between.

"PolyBase only works within PDW for now, but later it might be added to SQL Server (but there are no plans for that). PolyBase relies on the Data Movement Service (DMS) in PDW, and DMS does not exist in SQL Server"
http://www.jamesserra.com/archive/2014/02/polybase-explained/

Apparently things change. For those companies having file shares filled with structured, semi-structured, and unstructured data, replicating that data to a Hadoop HDFS cluster can allow you to immediately join it to SQL Server without ETL.

Or so I hear... :)

This might be a game-changer for some organizations.

Andrew Peterson spent some time with Polybase on his blog.
http://realizeddesign.blogspot.ca/

A few years ago I built a solution that utilized nothing but Analysis Services views to access data inside an Analysis Services multidimensional cube solution. Some of those views actually queried Excel spreadsheets, Access Databases and CSV files instead of SQL Server. At the time I thought it was kind of a mashup hack, though cool nonetheless due to lack of ETL effort and instant data availability.

Glad to see MS is catching up with this idea of schema-on-read, query-anywhere, anything. You're still going to need to create external, structured views on each file format to a specific Hadoop cluster, which seems rather linked-server'y. There are some other limitations to be aware of. You'll also need access to a Hadoop cluster of course. You do have one of those, right?

The Hybrid BI story is just beginning...
http://sqlmag.com/blog/what-coming-sql-server-2016-business-intelligence

SQL Treeo adds folders to your SSMS

2015-09-09T18:55:00.001-07:00

Microsoft, why can't you have packages like Oracle or at least a folder structure in Object Explorer?

SQL Treeo brings sanity to monolithic databases and multi-tenant servers, capturing objects into folders.

http://www.sqltreeo.com/wp/

SQL Server Collation Settings, Quelquechose de Vraiment Grav

2015-08-08T22:27:00.000-07:00

Saw this post about a Windows 10 error message a couple weeks ago on Facebook and laughed my head off.

http://bgr.com/2015/07/30/windows-10-something-happened-error-message/

Microsoft has gone ahead and adopted the Sad Mac philosophy of error handling.
From a consumer's perspective, I guess this yodel is a tad better than Catastrophic Failure and reassures you that something might happen next.

From a developers perspective, you really should have included the bugcheck code, stack trace, or at least some other kind of error identifying the method call you are finally giving up on.

If not, one fix for fellow Canadians installing Windows 10 will be to ensure you meld with our neighbours in the south by selecting English (US) as the regional locale. You selected French (Canada), didn't you?

http://www.technewstoday.com/25375-how-to-fix-something-happened-error-during-microsoft-windows-10-installatio/

Zut alors. Didn't SQL have this issue a few years ago? And Analysis Services?

Sorry MS, as a Windows 10 Insider I did not catch this bug. Actually, I didn't really catch any bugs. My Windows 10 install over the last year was a pleasant surprise, and I expect my main desktop to be upgraded soon. Perhaps my only suggestion would be that if English (Canada) is selected, then error messages should be bilingual. Seems like a waste of the second line of Something happened.

Instead of upgrading him, a family member who I provide tech support for asked me to just hide the "Upgrade to Windows 10" icon for him instead. From a support perspective I though that was wise for now.... I used the scheduled tasks and registry edit methods.

http://www.tenforums.com/tutorials/6596-get-windows-10-icon-remove-taskbar-windows-7-8-1-a.html

A few times in the last year I've encountered various collation issues related to Windows Regional Settings and SQL Server installs, and the defaults MS uses for products like Sharepoint and Reporting Services.

To switch over collation of a single database, you can try this at your own risk.

1. ALTER the database to change the collation (after killing all processes that are using the database and dropping all user defined functions with varchar max that bind to the collation)

2. ALTER the columns (this script may need to be modified to include schema id's when dealing with multiple schemas. This script was tested on SQL 2008 R2.

SELECT 'ALTER TABLE ' + o.NAME + ' ALTER COLUMN '
       + c.NAME + ' ' + t.NAME + '('
       + Cast(c.max_length AS NVARCHAR)
       + ') COLLATE ' + ' Latin1_General_CI_AS ' + CASE WHEN c.is_nullable = 1
       THEN
       'NULL' ELSE 'NOT NULL' END
       + ';'
--select o.name, c.name, t.name, c.collation_name, c.is_nullable, t.name, c.max_length
FROM   sys.objects o
       INNER JOIN sys.columns c
               ON c.object_id = o.object_id
       INNER JOIN sys.types t
               ON t.user_type_id = c.user_type_id
WHERE  c.collation_name IS NOT NULL
       AND o.type = 'u'

Ninite - mass installer

2015-07-29T19:28:00.000-07:00

https://ninite.com/ is a pretty awesome installer to build up a new PC. Fresh install of Windows 10? Check off Classic Start Menu and whatever optional software you want...

Ninite will

start working as soon as you run it
not bother you with any choices or options
install apps in their default location
say no to toolbars or extra junk
install 64-bit apps on 64-bit machines
install apps in your PC's language or one you choose
do all its work in the background
install the latest stable version of an app
skip up-to-date apps
skip any reboot requests from installers
use your proxy settings from Internet Explorer
download apps from each publisher's official site
verify digital signatures or hashes before running anything
work best if you turn off any web filters or firewalls
save you a lot of time!

Kerberos & Web Site

2015-02-11T19:34:00.002-08:00

Sometimes all you need is one tool for a tough job. If you're setting up Kerberos, that tool is the Kerberos Delegation Config & Report.

http://www.iis.net/downloads/community/2009/06/delegconfig-v2-beta-(delegation-kerberos-configuration-tool)

http://blogs.iis.net/brian-murphy-booth/archive/2007/03/09/the-biggest-mistake-serviceprincipalname-s.aspx

DOS tricks & tips

2014-11-03T18:37:00.000-08:00

Because sometimes simplicity is best.

type *.csv >> data.txt - consolidate multiple files to a single file

copy a.csv + b.csv c.csv - consolidate multiple files to a single file

http://www.powerpivotpro.com/2012/09/combine-multiple-worksheetsworkbooks-into-a-single-powerpivot-table/


Date formatting

The Myth of the Full-stack DBA

2014-07-04T06:34:00.002-07:00

Andy Shora recently published a great article about "The Myth of the Full-stack Developer." From his posts and tweets, I peg Andy as a front-end developer specializing in CSS and Javascript frameworks.

Andy's comment about the 30-point spider diagram really resonated with me.

Where do your web development skills lie?

The term Full-Stack applies to knowing different layers of architecture, of which the database (SQL/NoSQL/SomeSQL) can sit as one of those layers. However, if you were consider yourself a "Full-Stack" SQL Server DBA, you might know the details of some or more of the following components of SQL:

SQL Server

Transact-SQL
Backup/Restore
Replication
Log Shipping
Query Tuning and Indexing
Policies and Resource Governor
High availability
In Memory Querying

SQL Agent

Jobs
Alerts

Reporting Services

Sharepoint Integrated Mode
Native Mode

Integration Services

Importing/Exporting/Transforming Various Data Sources
Connection Types
Toolbox Components
Project /Package Deployment Methods

Analysis Services

MDX/DMX/DAX
Tabular
Multidimensional
Sharepoint

Powershell Scripting
TFS Integration

And of course the various ways to setup, configure, debug, maintain, monitor, license, upgrade and secure above tools along with the associated hardware and software. Perhaps your job description may even include mentoring developers on best practices on interfacing with the layer that is the "database".

By no means an exhaustive list and depending on your environment your role as a DBA could include Active Directory maintenance and Voicemail support.

If you're a Microsoft consultant or a partner of Microsoft, you're probably intimately familiar with the old Microsoft Hamburger and its associated toppings.

(RIP standalone PerformancePoint Server)

If you have expertise in one or more of the grills at the bottom of the stack, (SAP, Oracle, etc) your job just got a bit more important.

For some of those components in the stack, hopefully you're using best-of-breed third party tools from companies like Apex, Idera and Red-Gate to improve your experience. The ecosystem of helper apps is vast and wide. Many feature limitations of products are considered "partner opportunities" by software companies. If you want to be a Full-Stack DBA, you'll want to know what tools keep your stack upright.

If you were to consider yourself an expert on the internals of SQL, you might know about all of the components of this diagram, and take great pleasure in using tools like Internals Viewer to see how SQL arranges its storage and understanding the details of how GAM, SGAM & PFS Pages work.

Or you may have had a "crash" course on above internals while troubleshooting why a DBCC CheckDB job failed and your customers can't access their data. It's best to be proactive about these things, start studying the diagram!

The ScaryDBA's recent post about The Curse Of Relational Databases also comments about The Stack of Information Technology in general, and how deep, wide, and unfathomable it is. The comments provide insight into the mindset of the community, and one comment mentions the idea of "Technical Debt". A topic perhaps worth exploring into detail about in a future post, since this one is getting a tad rambling and epic.

Yesterday I attended a Toronto Hadoop User Group Meetup, with a discussion from the co-author of a book on YARN. The topic was a deep-dive into the internals and features of the resource scheduler and application manager that is YARN. Fascinating stuff (I'm not kidding!) that shows how much effort goes into the foundation of a scalable distributed data solution.

Speaking of deep, wide, and unfathomable....

The Big Data Developer Toolkit

http://blog.andreamostosi.name/big-data/

The Apache Open-Source Developer Toolkit

https://projects.apache.org/indexes/quick.html

Hopefully your DBA stack expertise includes some Linux knowledge.

Choosing a model - Tabular vs. Multidimensional

2014-06-23T05:35:00.001-07:00

Understand the limitations of the Tabular model and when Multidimensional model implementation is more appropriate.

It isn't as simple as Distinct Count = Tabular, Parent-Child = Multidimensional...
http://msdn.microsoft.com/en-us/library/hh994774.aspx
http://msdn.microsoft.com/en-us/library/hh212940.aspx

If you're still using Office 2007 or SQL 2008 R2, it might be a bit easier to make a choice...

Turing SQL Server - Is your power model sucking performance?

2014-06-17T03:06:00.003-07:00

As stated in this hilarious Knowledge Base article, changing your power options to “High Performance” (an option Microsoft conveniently hides under a “more” menu to prevent you from accidentally finding it) resolves this issue by using more power all the time. In return, you need to accept that global warming is entirely your fault. Amazingly, no performance analyser of SQL server suggests you change this option.

Sometimes the simplest things can drastically change performance. There should be an interview-style wizard or report for those who are new to setting up a server that would interrogate WMI to find out which settings may be sub-optimal for your usage scenario.

http://www.masterofmalt.com/software-development/blog/why-microsoft-windows-server-2012-produces-slow-and-inconsistent-sql-query-speeds/

cURL Super Powers

2014-06-17T03:02:00.002-07:00

cURL is an extremely powerful tool that you can use, available on many platforms including Linux or Windows (via tools like Cygwin). It's powered by libcurl. Wget is another like-minded utility to download files recursively.

Here's an example to parallel download all 150+ Microsoft eBooks available from this site.

curl http://blogs.msdn.com/b/mssmallbiz/archive/2013/06/28/almost-150-free-microsoft-ebooks-covering-windows-7-windows-8-office-2010-office-2013-office-365-office-web-apps-windows-server-2012-windows-phone-7-windows-phone-8-sql-server-2008-sql-server-2012-sharepoint-server-2010-s.aspx | grep -o -E "http://ligman.me/[0-9A-Za-z]+" | uniq | xargs -P 5 -n 1 wget --content-disposition

So what's it doing?

First, download (or curl) the url specified.
Second,search the second URL for a matching regular expression.
Third, uniq(ue) the resultset.
Fourth, xargs to deal with the large parameter list
Fifth, wget to download the resulting URLs in a parallel fashion.

Why use both tools when they appear to do similar things?

http://stackoverflow.com/questions/636339/what-is-better-curl-or-wget

If you're a Windows user, and can't install Cygwin, Powershell can do some of the things that wget can do.

http://superuser.com/questions/25538/what-is-the-windows-equivalent-of-wget

(new-object System.Net.WebClient).DownloadFile('http://www.xyz.net/file.txt','C:\tmp\file.tx‌t')

Here's a ton of GNU manuals, including wget.

Analysis Services 2012 and 2014 Performance Guide

2014-04-02T06:55:00.001-07:00

Microsoft has released an Analysis Services 2012 and 2014 Performance Guide, 111 pages of multidimensional performance tuning goodness.

Is there a future for the multidimensional cube at Microsoft? Sounds like it may stick around for awhile longer.... watch for improvements in SQL vNext.

Performance is important, and not just for improving business processes. Licensing rules for SQL 2014 will probably cause server administrators to try and keep their server utilization at full capacity. Do more with less.

A similar whitepaper exists for PowerPivot and the tabular model in SQL 2012.

SQL 2014 Released, so what's the big deal?

2014-04-01T12:06:00.001-07:00

SQL 2014 is released, in 2014 no less! You can download the bits from Microsoft or get your free Azure trial (with a cap to ensure you don't get charged anything) to try it out. Perhaps SQL Spade will help you to install it. Or just browse the new features on Virtual Labs.

Microsoft has officially killed support for SQL 2005 compatibility mode, so before upgrading from prior versions be sure you've upgraded the compatibility mode of your databases.

The database engine has various improvements, fixes and tweaks to query plan optimization and partitioning. Hopefully the query engine is tuned for the better.

There are some additional options for impersonation and separation of duties within the database.

Reporting Services, Analysis Services, and Integration Services, Data Quality Services, Master Data Management Services are mostly left in the dust. Sigh....

Some of the cloud integration stories have improved dramatically. Backup to URL. Migrate to VM. Hybrid blob storage. AlwaysOn to Azure. Azure's ExpressRoute functionality to move the public cloud to a private direct connection is now available. This should improve adoption for those not wanting to transfer data over the public internet. And provide companies with a secure form of offsite backup, DRP and replication to Azure.

Perhaps the biggest story with SQL 2014 is its In-Memory (formerly Hekaton) engine. Or what I like to call DBCC PINTABLE+++. This, along with its ability to use solid state storage for the SQL Buffer Pool, should provide for much faster query times when it's used correctly. Think columnstore indexes but for an entire table, and updateable, and not a separately stored object. Columnstore indexes in SQL 2014 can now be created as updateable, clustered objects that contain the data itself, in addition to the indexes. In Enterprise Edition, anyway...

Note for SQL Server 2005:
This (DBCC Pintable) functionality was introduced for performance in SQL Server version 6.5. DBCC PINTABLE has highly unwanted side-effects. These include the potential to damage the buffer pool. DBCC PINTABLE is not required and has been removed to prevent additional problems. The syntax for this command still works but does not affect the server.

Ahhh, memories of data corruption, and arguments with other technology experts that there is no such feature post-SQL 2000 for pinning tables in memory. "But I typed DBCC Pintable in and it came back with no errors!" Whenever I want to do something that bypasses the internal workings of a database, I like to visit Ask Tom. Even if you're not using Oracle, his site is a very fun an informative read.

Now back to SQL and the fine print.

Some of the licensing challenges to be aware of. One challenge, where can you get some budget for 128 GB of memory for your SQL Server Standard boxes, and 512GB of flash memory for your buffer pool? How about 4 terabytes on Datacentre with Enterprise edition, and some of those larger tables pinned in memory?

Fun times ahead with in-memory databases.

Compare and contrast user dynamically selected sub reports against a main report

2014-03-31T08:06:00.001-07:00

Provides some walkthroughs and samples of Reporting Services scenarios for SQL & MDX reports.

http://innovativebusinessintelligence.com.dnnmax.com/HCIEpisodes/DataSetQueryingonUserSelectedReports/tabid/95/Default.aspx

OCD Programming

2014-03-28T12:57:00.001-07:00

A great StackOverflow answer using the railroad analogy to explain Branch Processing and why processors love predictability.

http://stackoverflow.com/questions/11227809/why-is-processing-a-sorted-array-faster-than-an-unsorted-array

I think the response that won the votes was because of the picture in the response, and the answer "Consider a railroad junction" which succinctly describes the problem you may have when dealing with unsorted data and path analysis.

I am consistently amazed at how much difference a good path makes in programming, especially with databases. Ensure your tables all have primary keys. Performance improves. Add covering indexes for long running queries. Performance improves.

These two things can sometimes fix many of the performance issues living within the database layer.

Clean up your code. Ensure state is kept to a minimum, and conversations with slower systems such as databases and disk are batched up and cached.

And apparently sort anything coming down the pipe before trying to analyze it.

Predictability and patterns in your code, and being obsessive-compulsive when it comes to clean coding seems to mean a lot. To some compilers, anyway.

Others will ignore all your hard work and try to determine the best way to solve your problem. Seems to be the norm for the Oracle instance I'm currently using.

If you use a tool such as Windows Sysinternals process explorer, you can see what kind of effect these types of incorrect paths have on system performance. You'll notice a lot of Access Denied, Key not Found or File not Found messages. Systems could be so much faster if it just found the correct registry entry, filename, or internet address the first time, without going down multiple paths blindly. Perhaps predictive analytics will soon solve this issue, so that each time you use your computer it will intelligently speed up your daily tasks.

Or perhaps I'm just being obsessive.

Modern IE

2014-02-25T07:40:00.002-08:00

Great site from MS to download VMs for IE testing.

http://modern.ie/en-us

As a side effect, you can get VMs for Linux & Mac OS of all major MS operating systems!

Windows 2012 Command Prompt

2014-02-06T06:43:00.002-08:00

Gates Spends Entire First Day Back in Office Trying to Install Windows 8.1 : The New Yorker:

The New Yorker should have done their fact-checking. From what I have read Bill Gates is known for his trucker mouth when dealing with staff.

I downloaded 4GB VM of Windows Server 2012 trial yesterday to brush up on my Microsoft O/S skills and missed the fact it was the "core" edition. I was back in 1989 heaven when the DOS prompt came up on startup, after a nice UI for the login prompt.

Took me a few searches to figure out the UI wasn't broken, and some hacker-like Powershell to install the UI components. Ended up failing since I guess I need to download the install ISO too.

I wonder what Bill Gates would think when his 31-year history of DOS operating systems did a full circle with Win 2012 Server. Who would have thought in 2014 I would be staring at a command prompt typing "help".

'via Blog this'