Reed Me : database design

Nulls are always a problem.

reedme — Wed, 11 Mar 2009 20:33:11 GMT

Anybody who’s ever jousted with nullable database columns will find this humorous:

The Vermillion Screen of Death (Raw Materials Comic Strip)

Well, I did anyway.

Sybase loves AdventureWorks2008?

reedme — Thu, 22 Jan 2009 19:24:38 GMT

A SQL Ranger friend of mine in the UK gave me the heads up to this yesterday. (Thanks, Ben!!) Apparently, back in December 2008, Sybase ported some of our samples to SQL Anywhere. Needless to say, I’ve got mixed feelings about Sybase porting AdventureWorks to their platform. Sort of like a cherry-cheese Danish: it’s a pastry and pastries are supposed to taste good, right? But cherries and cheese? Ick. (I’m on a severely calorically restricted diet; sorry for the digression.)

As SQL Server 2008 and SQL Anywhere 11 are two very different database solutions, the ‘AdventureWorks2008’ database will appear slightly different, and offer a few functional differences. These incongruities are the result of different feature sets, which are intrinsic to each database’s design. AdventureWorks2008 Sample Database for SQL Anywhere Download: Mobile Enterprise, Database Management iAnywhere - Sybase Inc

We license AdventureWorks under the Ms-PL, so anybody’s welcome to do whatever they want with our code. It’s out there for learning purposes. Sybase seems to have learned that SQL Anywhere 11 doesn’t do a lot of the really cool things that SQL Server 2008 does! They only ported the OLTP database – I wonder why they ignored DW and LT??

The Sybase port of AdventureWorks2008 drops the following features:

They dropped all the spatial columns because SQL Anywhere doesn’t have a geography type.
They dropped Try/Catch error handling in all stored procedures, since they can’t do that.
They dropped DDL triggers, because SQL Anywhere doesn’t have them.
They dropped all stored procedures with recursive queries, because they don’t support them.
They dropped all Full Text indexes and queries, since SQL Anywhere doesn’t do that, either, I guess.
They dropped or modified all tables with HierarchyID columns, because they don’t have that data type.
They dropped all the tables with XML data, or omitted the XML columns, because they don’t support the same XML query lingo and have no XML index support.
They dropped all the indexed views, because SQL Anywhere doesn’t do that,either.
They dropped all the views that use PIVOT, because (guess what?) they don’t have that keyword.
They had to modify a bunch of stored procedures because they don’t support EXECUTE AS CALLER. (Did I mention that they had to drop all the Try/Catch support, too?)
They had to emasculate most of the trigger functionality (not counting the DDL triggers they had to drop outright), because they don’t support those features. Such as dropping Try/Catch support to throw exceptions instead of logging and always updating certain columns (even when not appropriate for the business rules):

HumanResources.dEmployee
Person.iuPerson
Production.iWorker
Production.uWorkOrder
Purchasing.iPurchaseOrderDetail
Purchasing.uPurchaseOrderHeader
Purchasing.uPurchaseOrderDetail
Purchasing.uPurchaseOrderHeader
Purchasing.dVendor
Sales.iduSalesOrderDetail
Sales.uSalesOrderHeader

They had to truncate all the binary image columns to 32,767 bytes (which will break all the ones bigger than that) because SQL Anywhere doesn’t have a large binary type apparently. Whoops!

Anyway, I don’t have a SQL Anywhere installation to try it out on, but that list from their own website reads (to me) like a list of reasons to use SQL Server 2008 instead. It looks like they ported our .NET samples for AdventureWorks2008, too, so I have to wonder if they’re equally limited as the AdventureWorks2008 port?

I do hope that SQL Anywhere customers do get some benefit out of using the port of our sample OLTP database (or switch to SQL Server 2008 where they can get the full functionality of AdventureWorks2008). I’d also like to invite Oracle and DB2 to port AdventureWorks2008 to their platforms so that the world can see how much of our feature set they don’t support, either. Heh.

AdventureWorks2010, anyone?

reedme — Fri, 31 Oct 2008 00:10:31 GMT

We’re doing planning work for SQL Server Kilimanjaro and the next SQL Server wave that follows it, which means product samples, community projects and AdventureWorks. Some people have suggested doing away with AdventureWorks in favor of something new, but I don’t see that happening... unless there is an enormous groundswell from the community on the subject. You decide.

As it is, about the only thing that we’ve “settled on” so far is that the schema will be based on AdventureWorks2008 and the working name of the database will be AdventureWorks2010. [Maybe it shows my age, but 2010 still feels like it’s a long way off… not just a short year and couple months.] Yeah, we’re decoupling a bit from the SQL Server release vehicle(s) so that we can rev AdventureWorks as needed instead of only at specific release milestones.

Anyway, we’re committed to being more open about our plans for AdventureWorks and our other open source projects. So come see what we’re working on, vote on the work items that are there, propose new work items if there’s something else that you’d like to see us do with the samples databases.

AdventureWorks Issue Tracker

Over the coming weeks, we’re planning to move all of our day2day source control management for all SQL Server product samples and community projects to CodePlex, so look for more activity and more regular activity than you’ve seen in the past.

I’ve had data days like that...

reedme — Sat, 27 Sep 2008 01:22:21 GMT

Brian Walker has a good pair of articles on Thou Shalt Nots™:

I had the chance to examine three business databases in five weeks and I'm still staggered by what I found. These are SQL Server databases serving as critical business foundations for companies generating millions of dollars in revenue per year. Every day, hundreds of employees are counting on these databases to deliver accuracy, stability and performance. Personally, I would not trust these databases to store a simple catalog of the 2,600 songs on my iPod. SQL Server database design disasters: What not to do (Part 1)

If any part of his article makes no sense to you, please do NOT participate in data modeling activities. Really. For the good of the children. Just don’t.

If he’s describing YOUR database project, I just want to say: “You’re doing it wrong!”

Do not assume that a typical DBA is a database developer. They are not the same thing. Their work may overlap somewhat, but they need different skills. SQL Server database design disasters: How it all starts (Part 2)

If your database developers are ALSO your production DBAs: “You’re doing it wrong!”

[Thanks, Brian. You’ve helped me added at least two new questions to my list of things to ask future employers before I take a new job…]

FOTR, if anybody writes new, untested SQL queries against a production database to gather reporting data for management, or you don’t have a change management process for production, including a runbook describing normal operating and HA/DR procedures, etc, then: “You’re doing it wrong!”

Props to Steve Jones and the crew @ http://www.sqlservercentral.com/ for linking up Brian’s great articles, or it might’ve been even longer before I noticed them.

For a humorous analogy about “doing it wrong” that you could use this XKCD comic to create an “HR incident” or “career limiting move.” You’re on your own if you run afoul of the analogy police.

Breaking News! AdventureWorks abandoned for Katmai in favor of Northwind, pubs and...

reedme — Wed, 02 Apr 2008 01:01:45 GMT

In a striking change of direction, I'm happy to announce that AdventureWorks is being deprecated and abandoned for SQL Server 2008 samples in favor of long-time favorite databases of simple-minded data modelers: Northwind and pubs.

Also, I can gleefully announce, in the spirit of oversimplification, that work is underway on ports of the Oracle sample schemas named:

The creatively-named Oracle sample schema IX "might" present a challenge for SQL Server; however, we're going to give it a college try... for April Fools' Day.

Heh.

Seriously, though, one person out of billions of (conscious or unconscious) SQL Server users has formally asked us to [re]include Northwind and pubs in our Official Family of Sample Databases™:

Feedback: Include option for Northwind Database install with SQL Server 2005 and SQL Server 2008 (connect.microsoft.com)

Since nobody has voted for it (not even the original requestor), we can only assume that it was filed on Connect as an early April Fools' joke.

[For the record, I hate April Fools' Day, but I am trying to get into the spirit of the faux holiday.]

How do I use Feature X in SQL Server?

reedme — Tue, 29 Jan 2008 00:20:18 GMT

Buck's got a great point about how SQL Server's development themes are managed from the top down:

It isn't a matter of thinking up single features or changes to the product, it's a cohesive direction we all want to go.
Carpe Datum : Why on earth are you working on that?

Reading his blog always make me wish that I could translate the pun in the title of my blog into Latin, too. Heh.

However, when it comes down to actually using SQL Server's new features du jour (see how I get my exotic language points in, too?), it's all about the individual features and changes to the product. Which is where the product samples and community projects come in... We (collectively) have got to take the roadmap and figure out the simplest way to show people how to use individual features and changes to the product in 2008.

Here are of the new features that we're fixin' to show off in the AdventureWorks 2008 family of databases. (Yes, I just moved from Tejas to Redmond, so we're fixin' to. Deal wit' it.)

Support for Entity Framework
Integrated full text indexing support
New data types

New date/time types
ORDPATH
FILESTREAM
GEOGRAPHY

Backup compression

At least that's the current plan. We're working through some changes that we tentatively plan to ship in the CTP6/joint launch timeframe. Among other things, I'd like to figure out how to get transparent data encryption and some others in there, too. BUT some of the self-imposed constraints for the AdventureWorks family of databases are that they have to be easy to deploy... and easy to understand. I've got half a mission statement for the AdventureWorks databases drafted, but those are two of the core principles.

Easy to understand is the driver that prompted Bonnie last year to create the newest member of the AdventureWorks family of databases: AdventureWorks LT. We (before I was part of we) had received feedback from customers that there are some folks who are new to the database universe who struggle with a higher level of normalization that the AdventureWorks OLTP database demonstrates, so AdventureWorks LT was born. It's a subset of the AdventureWorks data in a greatly denormalized format that is easier for instructors to demonstrate simple T-SQL concepts with and easier for new students of database technology to grasp. But I digress... If you yearn for the simplicity of Northwind or pubs, check out AdventureWorks LT and let me know what you think.

If you've got a favorite Feature X that you'd like to see in the AdventureWorks 2008 schemas, now is the time to voice your opinion. Let's hear it!

[Obligatory Disclaimer: Like the Any Key of legend, the name Feature X is a convenient placeholder or wildcard term. There is not now, nor has there ever been, a part of SQL Server called Feature X. As far as you know.]

Big brother (or SkyNet) is coming... to a PC near you.

reedme — Thu, 17 Jan 2008 04:18:24 GMT

I loved the book 1984, even though it was almost 1984 by the time I read it. The fact that the date was around the corner and technology wasn't anywhere close kind of took the punch out of the story. I guess it didn't scare me the way it scared "previous generations" (to put it politely). Perhaps being able to fool a lie detector later in life clinched it for me... Any system can be beaten; you just have to figure out how.

Labor unions said they fear that employees could be dismissed on the basis of a computer's assessment of their physiological state.
Microsoft System May Monitor Workers' Brains, Bodies (FOXNews.com)

I'm not concerned about the automated dismissal so much as I am about where and how we're going to store all of that telemetry data. Wow! Imagine how much sensor data would have to be collected every polling period (1 second?), centralized and sanitized to make it useful... Multiply that times 100,000 employees over the course of a couple years and you've got a serious data warehousing problem. Not to mention the AI for alerting someone in HR that a pink slip cannon needs to be activated.

Watching Terminator: The Sarah Connor Chronicles this past two evenings did make me starting knoodling about the complexities of the database systems required for SkyNet's C3 systems (command, control and communications). Wouldn't that be a fun database to model on SQL Server 2008? Wish I had more free time!

Fiction is rapidly becoming science. Brought to you by Microsoft.

Nominate your SQL Server wins for the 2008 TDWI Best Practices Awards!

reedme — Thu, 10 Jan 2008 21:11:10 GMT

Vote early and often for your favorite implementations on your favorite platform! (Hint: your favorite platform is SQL Server.)

Have you completed a data warehouse or business intelligence implementation that you think other organizations could learn from? Would you like some recognition for your accomplishment? Are you a vendor with successful customers who deserve recognition?
2008 TDWI Best Practices Awards

Although I'm working up a proposal for my own SQL Server 2008-oriented contest, I would *love* to SQL Server sweep all of these categories!

Enterprise Data Warehousing
Customer Intelligence
Dashboards and Scorecards
Enterprise BI
Operational BI
Data Governance
Master Data Management
BI/DW on a Limited Budget
Predictive or Text Analytics
Radical Business Intelligence
Government and Non-Profit

I worked on many, many engagements with customers and partners over the last year or so [before I left the field for the product team] who (I think) should be solid winners in many of these categories. You know who you are!

It's been real and it's been fun... Now here's The Quiz.

reedme — Fri, 30 Nov 2007 08:11:19 GMT

This month (on top of all the other excitement), I've been the facilitator for the MVP Academy course "Designing Microsoft® SQL Server™ 2005 Databases." (Calling me the instructor would be like putting lipstick on a pig -- with apologies to the pig.)

If you're not an MVP or if you've ever wondered what goes on inside an MVP Academy course, here's The Quiz that my vic-, er, students were given to wrap up the one month course (five MSL eLearning courses rolled up into one collection).

Please reply to me PRIVATELY with your new understanding of the following in order to receive credit for completion of the course and qualify for the exam voucher. You'll note that there is not a single "right" answer to many of these questions. Please treat this as a structured interview survey: pretend that I'm the hiring manager and that you actually want the job. ;-) http://blogs.msdn.com/reedme/archive/2007/05/30/the-quiz-lives.aspx

Oh, and it's due Monday. It's open book, but please don't bother cutting-and-pasting your answers from the course materials (or each other). I will send Chuck Norris to visit cheaters. (After he's done campaigning for Huckabee.)

1. Compare and contrast in your own words (preferably no more than 250 of them) the conceptual modeling methodologies ORM, ER, NIAM and UML as discussed in the course materials and/or the course newsgroup.

2. In your opinion, what is the most frequent mistake made with regards to data modeling artifacts?

3. Describe the salient points of the "surrogate key versus natural key" jihad. (Surrogate key is often used in the course text as a synonym for artificial key.)

4. Compare (or contrast, if you wish) two different auditing techniques for tracking application data manipulation: centralized (aka simple) audit versus history.

5. Discuss the implications for non-repudiation with regards to simple audit.

6. Which new feature of SQL Server 2005 is best suited for preserving the integrity of a database during a data migration?

7. Give me one good reason why you would ever want to use Snapshot Isolation.

8. Which SQL Server tool is the most helpful in determining the appropriate clustered index? Describe (briefly) how best to use this tool with the least amount of impact on production systems? How would you validate the clustered index selection was appropriate?

9. Imagine a scenario in which a CLR stored procedure would the ideal solution and describe the requirements which would justify the additional development and maintenance cost.

10. What is the Windows service which enables and manages distributed transactions? Describe the limitations of and caveats for using this service, particularly with regards to a heterogeneous enterprise platform environment. What is one possible alternative to using distributed transactions?