So the question to ask is: Is OBIEE up to the task?

Much of the reasoning for this topic arises from the combination of the following factors:

• Long development times for building a data warehouse.
• Sometimes typical short lifespan of the developed product.
• Excellent as usual blog post series from Rittman Mead.  OK, this took 8 months to get published but, by all means, if you have the time, read that series instead of this post.  No fluff at Rittman Mead’s Blog!
• Ever changing database architecture landscape.  (Including never ending new ways of storing and crunching data.)
• The monumental amount of technical planning and provisioning needed to take a BI project to completion for a customer.
• This should be a fun exercise.

Our Database Schema

I am using Oracle’s Order Entry (OE) Schema in this post.  This schema is generally available for download and it’s modeled in the typical transactional model.  Sample data populate scripts are provided as well.  It has been around for more than a decade thou and even Oracle ships more than one version concurrently depending of the database version installed.  Personal, Standard and Enterprise Edition versions of Oracle Database currently ship with one version and Enterprise with Partition ships with a different version.  All versions I have seen, however, are compatible with each other.  I bet Oracle picks a subset of features to include in demos depending on database features installed.

Obtaining Our Schema

If you have access to your database server installation directory, the scripts for installing and populating the OE Schema would be here:

_db_install_loc_/…/product/_db_version_/server/demo/schema/

If you do not have access to the server where your database server is installed or if the version you installed does not have the scripts, you can search and find them online.  I used the scripts I found at the following locations:

1. http://st-curriculum.oracle.com/obe/jdev/obe11jdev/11/common/files/sample_schema_scripts.zip
2. http://download.oracle.com/otndocs/tech/ias/portal/files/RG/OBE/common/OBEConnection.htm In this second link, search for 8i version of the scripts.

Lastly, if you want to review this and the other demo schemas Oracle bundles with their database, get Oracle Data Modeler and go here:

http://www.oracle.com/technetwork/developer-tools/datamodeler/sample-models-scripts-224531.html

Note -  It is totally worth putting the time into this tool.  It slices bread in 3 different locales for multiple database vendors.

Schema Deployment

I had to combine the above listed script sets in order to populate all the tables in the order entry schema.  I used the scripts from the first link to create and populate most tables.  I had to use scripts contained in the second link to populate the Warehouses, Inventories and Product Descriptions Tables.  Another good sample schema (from Oracle demos) is the Sales History Schema (SH).
Anyhow, the challenges and steps described here can easily be done on a schema of your choosing.

Schema Rules Overview

The following rules define the transactional model by which we are bound in this exercise.

1. One warehouse can have many inventories (quantity of product(s)).
2. One product can be in many inventories.
3. One product can have many descriptions (For text translations).
4. One customer can have many orders.
5. One order can have many items.
6. One product can be in many order items.
   Imagine the case where a product is priced different between first and second item bought (Get 50% off second pair at Skechers).
7. One category can have many products.

Schema Diagram

Here we get a clear picture of the transactional model for our Order Entry Schema.  The relationships defined in this illustration were picked up straight from the scripts Oracle provides.  The schema has been left as is expecting necessary modeling to be done in metadata layer.

SQL Join Overview To Satisfy Rules

   1:  SELECT    a.*, b.*, c.*, d.*, e.*, f.*, g.*, h.* 
   2:  FROM      OE.warehouses a

   3:            LEFT JOIN OE.inventories b ON a.warehouse_id = b.warehouse_id 
   4:            INNER JOIN OE.product_information c on b.product_id = c.product_id

   5:            LEFT JOIN OE.categories d ON c.category_id = d.category_id 
   6:            INNER JOIN OE.order_items e ON c.product_id = e.product_id

   7:            INNER JOIN OE.orders f ON e.order_id = f.order_id 
   8:            INNER JOIN OE.customers g ON f.customer_id = g.customer_id

   9:            LEFT JOIN OE.product_descriptions h ON c.product_id = h.product_id;

Importing into Admin Tool, we can validate that the relationships enforced in the order entry schema are intact.  Any modeling we do is based on foreign key relationships available.  If any relationships are missing from the model (no referential integrity enforced), this would need to be fixed by declaring them in the metadata.  This would ensure that OBIEE picks up the proper relationships for reporting.

Transactional Database in Admin Tool’s Physical Layer

This detailed view displays all the available columns for possible reporting.

After doing your aliases (or not) and ensuring all primary and foreign keys are declared, its time to move over to the Business Modeling and Mapping Layer (BMM).

For this order entry schema, I decided to design two distinct star schemas, one for Inventory and one for Ordering.  From here on, design decisions would need to be made to satisfy the project being worked on.  For this post, I just wanted to go thru the motions of creating a dimensional model on top of a transactional schema, hence the two star schemas devised.

The first Admin Tool feature to use is the ability to create logical tables from multiple physical tables as shown in the illustrations that follow.

Inventory Star Schema (BMM)

For Inventory Star Schema, I create a logical fact table from all the tables that previously validated (and defined) foreign keys allow.  These relationships basically defined (logically) this table for our model.  Notice that Product Description Table is not included in this logical table.  If model required the fact able to hold the surrogate key for the translated text, it would have been necessary to declare that relationship before this step.

From here, modeling the dimension tables will follow the same, albeit simpler, pattern as described in following illustration.

Once all the desired dimensions are created n the BMM, we end up with our star schema.

Expanding all objects in schema shows all the elements available in reporting.  This effectively maintains an OLTP to OLAP layer in the metadata for us.  As we use these star schemas in reporting, the SQL generated by OBIEE will rely on these declarations to fetch from our relational system.

Ordering Star Schema (BMM)

The steps for the Ordering Star Schema are exactly the same. First we define our fact table.  All the necessary dimensional tables stem from the keys we can include here.  Again, the available keys are based on our primary/foreign key previously declared.

following the pattern used for Inventory, the end result is the Fact Ordering Schema (as Kimbal would have called) as displayed here.

It is important to note that, in order to enrich this schema, I unnecessarily modeled multiple dimensions that provide no extra value.  This is nothing more than a design exercise for this post.  Had this been a real project, it would’ve proved of little value to create dimensions for which we only have a key and key provides the only bit of information we have.  For example, the State, Country and City Dimensions.  In the background, OBIEE should just pick the key available in our fact table and not use them.  For this post, however, I desire a rich model to explore how flexible OBIEE reporting is.

Having completed our modeling, the last step is to expose our schemas into their respective subject areas for reporting.

Conclusion

It would be interesting to do A/B testing between reporting against this transactional model versus a ‘proper’ dimensional model in the physical layer.  Given identical hardware, it would be critical to define the point of diminishing returns in skipping the data warehouse all together.  Then again, whatever the findings are, it would be worth a complete lab to establish what can be done one way or the other in transactional versus dimensional reporting in OBIEE.  These tool capabilities, together with different storage technologies and even Oracle latest’s acquisitions should provide multiple reporting options to fit any case in the wild.  Maybe I can explore some of these in future posts.

As always, thanks for reading!

You may know about the site DC Rainmaker, what I think to be one of the best run/swim/bike blogs that I know.  I find Ray Maker’s Run-Around-The-Neighborhood posts very entertaining.  As of this writing, his latest is A London Runaround.  Go check it out, he is very enthusiastic and you can’t help but enjoy his runs. 

Paying homage to these, I decided to find out if these are as fun to do as they look!

May in Orlando means we try to schedule our runs early in the morning or late in the evening.  This morning before work, I decided to head out for a morning run. 

Sidenote – If you’re in Orlando for business, this run is close to the airport.

I start my run imagining I am cruising down Beverly Hills.

Which quickly turns into Anytown USA, your typical office park with a mixture of businesses and hotels.

About 10 minutes into run though, the landscape is completely different.

Before long, we run into the littlest city close to Orlando Airport no one has heard about, Belle Isle.

Pretty soon, we get to one of many public parks around the city, Warren Park. 

Not many people here yet this early in the morning.

Running past, we see quaint shops and restaurants dot the route.

They even have their own gas station with a waterfront.  Neat. 

With a little dock that allows people to enjoy Lake Conway.

Along the route, most streets leading to the lake have No Outlet signs posted. 

A particular one, however, holds a secret for runners.  Note the fence at the end of the road.

It leads into the second park I am going to visit today, Lagoon Park.

Wedged between neighbors, this is a very tranquil place to take a breather and sip some water.

Its Tuesday so rest has to wait. I head out to the neighborhood on the other side of the park.

And get to enjoy a nice run with a lot of shade.

Soon, I stop at a not-quite-a-park but a boat ramp.  It is a big lake after all.  This is called the Perkins Boat Ramp.

Where we can enjoy the nice view.

And greet the neighbors as they go about their day.

Not a mile goes by before I visit the third official park of the day, Lucille B. Bateman Park at Swann Beach.

Heading back to run; I almost forgot I was in long run.

I pass a beautiful house I could live in but don’t.

As I head out to cross Lake Conway on the way back to work.

This is a lot better place to drink take a sip of water with a nice view at either side.  Running in the morning makes all things peaceful (except dogs).

I can almost picture myself in a kayak instead.

Back to reality, I cross the bridge and head back the other way.

Looking back assuring myself I’ll come back.

Forget the kayak, I could buy that boat instead.

Here is the last park, I think fourth, I will visit in run today.  It is the smallest park I’ve ever seen, Gilbert Park.

Nice little nook. I wonder if it provides public access to the lake.  I’ll find some other time.

So I actually run into the last park, fifth, of my run today.  It is actually part of  Warren Park, the first park I ran into today. This part is on the other side of the road so I count separately.  This would be a nice place to come play with the kids at lunch break.  I’ll have to ask my wife.

It even has the requisite playground for the kids.

Exiting back the way I came, I see people enjoying the park by now.

Sadly, I have to head back for a full day of work ahead.

Bye Belle Isle.

Sorry for grumpy face. A quick shower and a full day of work awaits me.

Not as picturesque as a run around London but it sure beats sitting at desk.  Lets wrap this up with required route and link of morning run. 

As Ray would say, thanks for reading!

With this in mind and looking for an excuse to keep playing with Awesome Tableau Software, I decided to take another look at the Healthcare Costs Data previously talked about at Centers for Medicare and Medicaid Services’.

This time, I restricted the data to the state of Florida (I’m here) and decided to drop the map.  The hopeful intent of this visualization is to provide the needed information as fast as possible.  As before, I cannot embed here but provide the link instead.  Just click on the images to play with visual.

And for perspective, this is what Stephen did.

Click on the image for visualization.  Free WordPress won’t let me embed!

During a run, you can scroll thru these two screens, plus time and date screen, using the lower right button on the watch. In this way, the watch can display up to 4 data fields while you run.

On a different setting option, you can set Pace to be average pace or instantaneous pace (default). This setting is absolute, you cannot have 2 pace data fields, one for average and another one for instantaneous.

Having set the two display screens already, including (now average) pace, you have the option to include instantaneous pace as well.

Just turn on Virtual Pace. This feature takes a pace given and informs you if you are ahead or behind it.

Now, as you run, you can scroll thru these screens with the Virtual Pace screen now displaying a fifth data field.

Sweet!

Data looks like this:

<rant>

One of the things that most attracted me to the field of BI was the opportunity to add value to any situation.  Data analysis can positively leverage any situation.  In the absence of information, choices we have to make are not always clear.  In a nutshell, that is why I gravitated towards this field.

However, I haven’t found many opportunities ‘greater good’ opportunities professionally.  I hope actions like this start snowballing.  This is indeed good stuff.  This data should be free!

</rant>

I had registered to run my first half marathon, Lake Minneola’s Half Marathon.

Today was not going to be about following my first doctor’s advice to lose 30 pounds.  It wasn’t going to be about my second doctor’s advice to lose 30 pounds either. It was not going to be about recurring back pains. Today was only going to be about running. Five months of telling myself speed and distance are not the point be darned. People express feeling alive when they are having fun.  Often, fun comes when one is out of their comfort zone. Running does that.

My precious wife had agreed to drive the hour from Orlando to Minneola for the race.  I notice her get up groggily with an epic task: get 2 kids up and ready for upcoming morning. While she did all the hard work, I tied my shoelaces 3 times and reviewed my music playlist for the run. I just have to run.

Before long, there’s two kids wrapped in blankets and a cooler full of snacks in the minivan. Kids had breakfast bowls in hands as well.

5:40 a.m. – My wife had already won her first race.  Wondering what was I supposed to eat, we set off to Lake Minneola’s Annual Half Marathon.  Where is my granola?

6:30 a.m. – Leaving the rain behind, we arrive at the Lake Minneola. The morning glow reveals an otherwise serene lake disturbed only by the incoming breeze from the lake.  Small groups of people are making their way to pick up their race packets.  Wisdom of the crowds, we join them. 

6:40 a.m. – I am putting on my bib on.  Still wrapped in blankets, the kids are snacking.  My wife notices that there is not much to do before a race.  I agree with her and wonder…  I wonder if this the time to worry about the race?  Runners around me do not seem to be as pumped as I’ve seen in shorter races.  No one seems worried either.

I had run OUCs Corporate 5K this past Thursday.  This race was more than 4 times as long, this race was going to be different.  I do long runs on Sundays and only a few of these were more than 13 miles.  None were under two and a half hours long either.  Under the advisement from fellow runners, I had even taken energy gel packs (GU) to take every five miles or so on these runs.  The main goal being not to run better but to asses whether my body would agree with them.  They felt like taking mouthfuls of condensed milk in a hurry.  Race nutrition: total win.  More worrisome from long runs, I often run out of water.  For this race, I intend to stop at every water station, no exceptions.  My goal is to finish the race, maybe beat all my Sunday runs.  I had rehearsed in my mind what and when was I going to refuel, to drink water and to drink Gatorade.  I had even pre planned a silent celebration at mile seven to signify past midpoint of race gone.

By now, the kids are awake and helping themselves to the runner’s fruits and water.  They look right at home so my wife and I relax for a bit taking in the sights.  The sun is coming out by now.  Lake Minneola is very pretty and not overly crowded with buildings.  Seems like a nice place to run.  The race is going to be mainly around the lake; 13 miles sure looks big right now.  The weather seems perfect for running thou.  It’s a cool 64^o and the wind is howling its way inland.  Nearby, small groups of people are training for triathlons.  They are keeping warm by pedaling in stationary bicycles wearing full body suits near the shore.  Brave souls indeed.

6:50 a.m. – I walk into pit of runners and try to place myself smack between front of line and middle of the pack.

7:00 a.m. – Race is on! Phone app is active and I am psyched.  Anytime now, the people in front are going to start walking away;  I am already planning which lines to take to clear them all.

7:05 a.m. – No one is moving.  This is not corporate 5k,  there’s not 16,500 runners here.  Why aren’t we moving?

7:15 a.m. – Seriously, we should be a few miles into race by now.  Why are there runners making small talk, some even made their way to lawn to get some space.

7:20 a.m. – Come on! We should be 1/4 of way into race by now.  I can sense the battery depleting on phone.  This is not cool guys.

7:25 a.m. – A guy walks down starting line blasting an air horn and screaming the race has started.  I wanted one of those starting line gun fire blast… This feels like the biggest casual run meet ever. 

7:30 a.m. – The walk begins.  I find out later there is less than 500 people running.  This would be considered a small race.  We do seem to be packed a bit tightly for the first mile or so.  For my next half marathon,  I am moving up closer to start-line.  Let them pass by me instead.  Its just that people looked so much more ‘racy’ up front. 

The running app on my  phone is telling me I am keeping a 9’30” pace.  I had planned to run a steady but slower pace than this.  I am positive this pace is too fast to keep for the duration of the race.  Spur of the moment, I reason I can keep this pace for first half of the race and enjoy this feeling a little bit longer.  I am certain that, by mile six or seven, the crowd would’ve thinned out by then.  Then it would be the proper time to revise my pace.  With my mind convinced, logic was out of my head and I was just enjoying my morning run.

The first third of the race was not around the big lake.  We are running around the adjacent neighborhood instead.  I guess all that fine scenery will have to wait.  The houses around are pretty enough and the sights were all new anyways.

Oh my, we are running on the road WITH cars!  It was the usual slow moving neighborhood traffic but this was unexpected for my first half… to be sharing the road with cars.  In all the (few) races I’ve participated in, the streets have always been closed to cars.  I can run with cars a lot closer to home for free.  Oh well, closer to home we have no hills like on this ‘fast and flat’ race.  We get back to lake soon enough only to detour again.  This time, we take on a small trail that is very nice and has no cars. We are sharing this part of the path with road bicycles instead.  Compared to running alongside cars, riders seem pretty tame and friendly. 

About 4 miles into half (no mile markers, running app will do), I am still keeping my bit-too-fast pace.  I feel fresh as ever but I am wondering not if but at which mile will I feel the burn set in.  Hopefully, I will make another mile or two before I take my gel pack.  I am reminding myself I am going to stop at every water station, alternating between Gatorade and water.  Mile five goes by and I drink my energy pack, still feeling fresh and keeping awesome pace!

8:30 a.m. – Having set running app to announce distance at every quarter mile, I was eager to listen to mile seven come up anytime now.  Suddenly, music fades out as it always does before mile announcement.  ’6.96 miles, average pace: 9’35” per mile’  I unconsciously raise my hands in triumph; its all downhill from here. 

Announcer:

9’35” per mile’, ’6.96 mile, average pace, 9’35” per mile’

9’35” per mile’, ’6.96 mile, average pace, 9’35” per mile’

9’35” per mile’, ’6.96 mile, average pace, 9’35” per mile’

I frantically, repeatedly, whack the phone while trying not to slow my pace down.  Coincidentally, I also hear two music tracks playing at the same time. I suspect this is the result of another music app turning on with all the bouncing in my fanny pack.  I can only assume this led to Nike App crapping out.  In what seems like minutes, I somehow manage to stop one music track while my mind rejoins the rest of my body in the race.

I keep running at my current pace (I think) and enjoy the views.  The lake is very picturesque and that gel packet must be heavenly because I feel as fresh as can be.  I suspect both the cool weather and the time of day have plenty to do with this feeling.  Maybe I should do Sunday runs this early.  I notice, however, that my announcer never comes back.  Plan B – I try to note the people around me hoping to keep up with fellow runners.  This feels even better than listening to app.  I’ve  even been steadily passing people for a while now.

Sticking to my hydration plan, I spot the next water station ahead.  Odd.  There are so many runners standing around.  Why would that be?  Celebratory punch?  Nope.  This is a self-service water station and I need to take a sip!  Reminding myself its all about feeling alive and finishing the race, I line up and wait my turn.  Peeved, I take two cups of water.  With no mile markers in sight and my announcer M.I.A., I estimate when to take second energy packet and the pace to keep for the remainder of the race.  Aside from crossing a little bridge alongside oncoming traffic, the rest of the run proves to be as fun as can be. I only wished there were mile markers closer to the finish line to know when to pick up pace. 

9:30 a.m. – In the end, I kept my pace for duration of the race.  I had just finished my first half marathon in 2:05:49 at an average 9:37 pace!

First medal in hand.  I am definitely doing this again.

I learned a few things from this race as well.  I learned not to focus so much on silly running app but to focus more on running instead.  Maybe I get a watch in the future   I found out as well that nutrition and hydration during a race make a significant difference in running fun.  Weather plays a part as well.  I had never set out to run at 7:00 a.m. before and I had never ran this distance this fast either.  I learned cotton shirts are not a runner’s friend, that blue and red make green and I learned that you get a new set after about four days anyways.  Maybe I’ll try to take pictures during race as well.

I am already wondering what race to join next!

FYI – Music playlist with order of play plan:

1. Electro Dance 2 – complete disk is perfect for keeping decent tempo up
2. Neil Young, Harvest Moon – because the middle of the run was going to be a time to conserve energy so have enough for last third of race
3. Sonique, Hear My Cry w/out slow songs – so that I could pick up pace considerable faster than middle but slower than first half

Due to phone issues described above and the fact that phone shuffled my songs, this listening plan fell flat.

Notes of value follow each picture.

Here’s finished shelf.  The stain has almost dried, can still smell it.  I blink, its half way full already!

Here is the walkthru in approximate order

First, we approximately mark the notches both on the 2x4s and on the horizontal pieces of wood.

To get here, I used a circular saw, two cuts, and a flat-head screwdriver as my chisel.

Before long, I ended up with notches, a lot of them.

My wife suggested I take a few dangerous corners out to save a few scrapes and bruises.  Good idea!

Pre-drilling the verticals in preparation of the wall studs.

And filling them notches hoping angles are good enough.  Better tools would ensure better precision!  No t as much fun thou.

This is the last time this wall looked so neat. So sad.

And this is point of no return.  We are doing this.  Worth sharing that neither the verticals nor the wall studs are straight!

Here, you can see me test fitting the horizontal pieces for measuring the notches to cut.

This is proof there is not a straight line in the whole project.

Seven years old is FUN.  I wish I was seven.

One coat of stain did this.  Pretty.

But then I had to do the rest.  Wife helped lots. Goodness this smells strong.  Had to leave in garage overnight waiting for the fumes to evaporate a bit.

And this is the finished shelf, shim and all.  I had to use shims in about half of the connections to ensure 90 degree angles for all shelves.

That’s it’ enjoy.

[nQSError: 15019] Table Calendar is functionally dependent upon level Month, but a more detailed child level has associated columns from that same table or a more detailed table.

Where It Happens

I encountered this error while trying to create Calendar Hierarchy as shown:

Reading the table name in question from our error message, we go to the Business Model & Mapping (BMM) and examine the logical Calendar table. 

Why It Happens

Consistency check is complaining about our logical Calendar Table having more than one key in BMM Layer.  This error seems counter intuitive to me as I all these keys depending which aggregation table I would like used depending on granularity desired. 

I am certain my questions will be answered as I continue working on this RPD.

Solution

Removing this error (and additional ones for all extra keys in Calendar Table) required removing keys until only lowest key was left.

To get the sought after

And allowing us to save our model.

With the release of OBIEE 11.1.1.6, Oracle now provides basic SVN integration in Admin Tool.  It seems Oracle is moving away from storing RPD content as a binary file (wishful thinking).  I think this is a step in the right direction and would like to be able to merge and diff RPD work from Admin Tool or versioning system of choice.  I am assuming SVN integration in Admin Tool is just as good as in SQL Developer Data Modeler .  Just like Data Modeler, you still have the option of not enabling SVN integration and using your versioning system of choice.  This is what I am going to do.  It gives me more choice such as which versioning system to use and independence from Oracle’s SVN feature set included in tool which is a subset of SVN anyways.

For now, versioning RPD files is a two-step process.  We still have to deploy binary RPD file.  The general idea is to convert RPD file into XML directory structure to put in your version control system of choice and convert back into binary to deploy.

Overview

Starting from your RPD file, go to File, then Copy As and save your work as MDS XML Documents…

Tool asks you to Select XML Directory to store the contents of RPD as XML.

Select your directory and press OK.

Instead of the RPD file, this directory is what we would put up in version control.  At this point, we would do all our work from this MDS XML Document directory.

Open this location now form Admin Tool.

Notice that now, Admin Tool states that we are working out of directory instead of RPD file

Inside the directory selected (or created), You end up with a directory structure of XML files.

And inside each directory, you end up with a list of XML files, one for each element of the directory in question.  I.E. – For the the Logical Table directory, you end up with an XML file per logical table in your model.

This is were it gets a bit tricky. There is no easy way to differentiate one of these XML files from the thousand of others in directory structure.  This will make diffing and merging this directory lots of labor.  Still, it is way better than MUD or multiple developers with inconsistent copies of an RPD file.

Since these files are XML, however, we can use our favorite editor or tool to, at least, inspect differences between iterations of our work.

I would be hesitant to do much editing of these files for fear of rendering them unusable by the Admin tool. At least we have backups now .

After you finish development (and pass review before committing), you will have to convert this XML directory back into RPD format to deploy.  OBIEE can only deploy (for now?) RPD.

That is all there is to it.  You end up with RPD representation of your, now versioned, model work.  Yay!