Slowly changing Facts - data-warehouse

We'he slowly changing dimension in DWH. Type 1, Type 2, Type 3 etc. This allows us to track changes in dimensions tables. But, what is the way to track changes in the fact table?

There are a number of different ways to track history in a fact table, and a similar range of techniques can be applied as those used in dimensions. As always the technique you choose depends on the types of changes you have coming through and the ways in which users will want to analyse the data.
For example:
You might be lucky and not need to track the history of a fact table at all, as with many transactional facts which are just inserts of new rows.
You could apply the equivalent of a Type 1 change to a Fact record, updating and replacing the values.
You might supplement your design with a periodic snapshot, to see the measures at a point in time or for a defined period.
It might be the history of a long running process might be best dealt with in an accumulating snapshot fact, for example an insurance claim which might have various dates (claim date, resolution date) and measures (claim amount, resolution amount) which only transpire over time. In this case you are filling in the fact over time and tracking history that way.
You might need to supplement your accumulating snapshot with time stamped row versions, similar to a Type 2 change.
You might present simple versions of this complex picture behind the scenes, for most users who don't care about history, with views.
Some of these techniques may benefit from having surrogate keys on the fact tables.
This article gives a few hints into this more complex world of tracking changes in facts:
Kimball Design Tip

Related

Transaction Fact Table approach

I'm working on financial data mart structure.
And I'm having some doubts on whats the better approach to do so.
The source system database,Dynamics AX 2009, has three tables for customer transaction.
One table for open transactions, where the Customer still needs to pay for service/product;
One table for settle transactions, where it holds what the customer have already paid;
Finally a table that have all customers transactions, holds transactions from open to settle and also others transactions as customer to bank or ledger accounts.
I thought in two options, first I will maintain a fact table representing the three table, fact for open transactions, fact for any customer transaction and fact for settle transaction.
Second is to create a single fact to hold all transactions, to do so I would have to do a full join on three tables.
I'm not sure on both approaches, as the first seems to copy tables from production and create the proper dimension.
On the Second one I would create a massive fact table, that where data would constantly change, as open transaction are delete on source system when they are settle.
Another doubt, should i create a fact with scd(slowly changing dimension) structure to maintain history data?(star date, end date , flag)
It's hard to say from the information given whether this needs to be one or more Fact tables. However, the key point which you should use to decide is whether all of the information is at the same granularity. Consider the grain of your intended Fact table(s) and you should find an answer for whether you need one table or multiple tables.
If all of the information sits at the same grain - i.e. all of the same dimensions apply to all of the measures you are considering putting into the same Fact table - then they can probably all live in the same Fact table. If you're finding that some of the Dimensions wouldn't apply to some of the measures then you probably need to re-think your design. Either you might need multiple Fact tables, or you might need to take all of your measures down to the lowest grain and combine hierarchies into single Dimensions if you currently have them split across multiple Dimensions.
While it's been mentioned that having measures in separate cubes could make it difficult to compare things, keep in mind that you don't need one cube per Fact table. You can have multiple Fact tables in a single cube, and sometimes this is very helpful when you need to be able to compare measures which share some Dimensions but not others. This is far, far better than forcing data which does not have the same grain into one Fact table.
Also, it sounds like what you're trying to model is the sales ledger of an organisation. I'd suggest having a dig around via Google as you may well be able to find materials discussing dimensional data warehouse design for sales ledger structures, rather than reinventing the wheel. If you don't have a decent understanding of the accounting concepts you're trying to model I would especially recommend looking for a reference schema to work from, or failing that doing some reading up on accountancy concepts (and sales ledgers specifically). Understanding the account structure should help you understand what the grain of your Fact table(s) needs to be, how to model the Dimensions, and so on.
This is a really helpful abridged version of Kimball's modelling techniques which discusses grain, and the different types of Fact table, amongst many other topics:
http://www.kimballgroup.com/wp-content/uploads/2013/08/2013.09-Kimball-Dimensional-Modeling-Techniques11.pdf
I think you should just use one fact table (one cube) and use a dimension to differentiate between open/settled/etc. transactions. That's what dimensions are for: They help you to categorize your measures and get a specific view on them. This approach will also open much more possibilities to create knowledge with your cube. With separate cubes for open/settled/etc. transactions, it will be harder or not possible to set this data into contrast.
Since the data is changing constantly, you should consider to update your fact table in a given time and rebuild your cube if it needs to.
If you use scd or not really depends on the data you process and what it is used for. Is there a business case claiming it? Is there a technical use?
I think this is something you have to decide on your own.

Natural Key and Fact tables

I'm new on dimensional modelling I believe that you guys can help me in the following doubts.
In the production system I have a transaction table, sales table for example.The unique identifier is a primary key called SaleId.
Example:
My doubt is when modelling the fact table should the SaleID be included in the fact table as a NaturalKey?
Also should the Fact table have a SurrogateKey?
Please feel free to send me any link as reference.
Thanks in advance
Technically speaking, it is probably not a natural key - it does look system generated. However, sometimes it is very valid to store a system generated ID in a Fact for use as a Degenerate Dimension. Usually, these are cases where either the business users do have sight of this system generated ID (order numbers, invoice numbers, purchase order numbers, etc.), or where there's no other useful way of identifying some rows which can be usefully grouped together.
If the users of your BI solutions are likely to want to be able to drill down into information and look at it by sale, then the SaleID might well be a good candidate for this treatment. Have a think whether there's any other way for them to get to this level - could a customer be associated with two distinct sales on the same day? If so, would your users want to look at them as two separate sales? You might need to speak to them to find out what's going to be useful for them. If for some reason you can't get a clear answer, I'd say keep it - there's little harm, and you can always remove it later if it's not used.
Here's the Kimball group's take on Degenerate Dimensions, in case you're at all unclear on how they work:
http://www.kimballgroup.com/2003/06/design-tip-46-another-look-at-degenerate-dimensions/
As far as Fact table surrogate keys - I always use them. As Kimball's Design Tip #81 points out, they're sometimes useful, and it's the kind of thing I'd rather put in at the beginning and not use than realise later on that it would have been useful to have. Point 2 - where you might want to make updates by inserting new rows and deleting the old ones - certainly applies to work I've done.
The requirement for a primary key in a fact table depends on the type of the fact table. Transactional facts which are never updated do not need it. Periodic snapshots probably don't need it, unless the current period is a to-date measure. Accumulating snapshots definitely need it.

Core Data Model Design - Attributes vs Entities

I've been developing a very basic core data application for over a year now (Toy Collector, http://bit.ly/tocapp), and I'm looking at doing a redesign so that I can build in iCloud support. I figured while I'm doing that, I might as well update my core data model (if needed), and I'm having a heck of a time tracking down "best practices" for the following:
Currently, I have 2 entities:
Toy, Keywords
Toy has all the information about the object: Name, Year, Set, imageName, Owned, Wanted, Manufacturer, etc, (18 attributes in all)
Keywords has the normalized words to help speed up the search
My question is whether or not there is any advantage to breaking out some of the Toy attributes into their own entities. For example, I could have a manufacturer entity that stores the dozen or so manufacturers, instead of keeping that information in the Toy object. My gut tells me this could reduce the memory footprint (instead of 50,000 objects storing a manufacturer string, there would simple be 12 manufacturer strings in an entity with a relationship to the main Toy entity). Does that kind of organization really matter? Am I trying to overcomplicate things? I just feel like my entity has a lot of attributes, and I'm not sure if taking the time to break it apart into multiple entities would make a difference.
Any advice or pointers would be appreciated!
Zack
Your question is pretty broad, since it addresses the topic of database design. Let me say upfront that it is almost impossible to give you any sensible suggestions, since I would need to know a lot more about your app, use cases, etc. than it is possible through a S.O. question.
Coming to your concrete questions, I would say that you correctly identify one of advantages of splitting a table into multiple ones; actually, the advantage of doing that is not just reducing the database footprint, rather keep data redundancy to a minimum. Redundancy not only affects memory footprint but also manageability and modifiability of your data, and lack of redundancy could even cause anomalies or corruption. There is even a whole database theory topic which is known as database normalisation that addresses this king of concerns.
On the other hand, as it is always the case, redundancy can help performance, and this is actually the case when you can fetch your data through a simple query instead of multiple queries or table joins. There is a technique to improving a database performance which is known as database denormalization and is the exact opposite to normalisation. Your current scheme is fully denormalized.
Using Core Data, which is a relational object graph manager running often on top of SQLite, which is a relational database manager, you have also to take into account the fact that Core Data will automatically build your object graph and fetch into memory the data when you need it. This means that if you can take a smaller memory footprint on disk for granted, this might not be the case when it comes to RAM footprint of your query results (Core Data will "explode", so to say, at some moment your data from multiple tables into one object plus its attributes).
In your specific case, you should also possibly take into account the cost of migrating your existing user base (if the database is not read-only).
All in all, I would say that if your app does not have any database footprint issues at the moment; if you do not feel that creating new tables might be useful, e.g., in order to add new functionality, such as listing all manufacturers; and, finally, if you do not foresee tasks like renaming a manufacturer or such at some point, then maybe refactoring your database will not add much benefit. But, as I say, without knowing your app in detail and your roadmap for it, it is difficult to say anything really on spot. In any case, I hope this general considerations will help you take a decision.
EDIT:
If you want to investigate your core data performance and try to understand where the bottlenecks are, give a try to Instruments/Core Data tool (Product/Profile menu). There are a lot of things that can go bad.
On the other hand, it is really hard to help you further without having more details about the type of searches your app allows to do. One thing that is not clear to me is if your searches are slow only when they return a lot of results or they are slow even when returning a few results.
Normalizing might help performance if you only use (say, after doing a search) just one normalized entity (e.g., to display the toy name in a table). In this case all of the attributes referring to other entities would be faults (hence would not occupy memory nor take) and this might speed up things. But, if you do a search and then display the information from the other tables as well, then there might not be any advantage, quite the opposite, since the faults would have to be resolved immediately and this would produce more accesses to the database.
Also it is true that depending on how you use it, core data could not be the best way to handle your data. Have a look at this Brent Simmons' post relating his experience.

What database should I use in an app where my models don't represent different ideas, but instead different types with overlapping fields?

I'm building an application where I will be gathering statistics from a game. Essentially, I will be parsing logs where each line is a game event. There are around 50 different kinds of events, but a lot of them are related. Each event has a specific set of values associated with it, and related events share a lot of these attributes. Overall there are around 50 attributes, but any given event only has around 5-10 attributes.
I would like to use Rails for the backend. Most of the queries will be event type related, meaning that I don't especially care about how two event types relate with each other in any given round, as much as I care about data from a single event type across many rounds. What kind of schema should I be building and what kind of database should I be using?
Given a relational database, I have thought of the following:
Have a flat structure, where there are only a couple of tables, but the events table has as many columns as there are overall event attributes. This would result in a lot of nulls in every row, but it would let me easily access what I need.
Have a table for each event type, among other things. This would let me save space and improve performance, but it seems excessive to have that many tables given that events aren't really seperate 'ideas'.
Group related events together, minimizing both the numbers of tables and number of attributes per table. The problem then becomes the grouping. It is far from clear cut, and it could take a long time to properly establish event supertypes. Also, it doesn't completely solve the problem of there being a fair amount of nils.
It was also suggested that I look into using a NoSQL database, such as MongoDB. It seems very applicable in this case, but I've never used a non-relational database before. It seems like I would still need a lot of different models, even though I wouldn't have tables for each one.
Any ideas?
This feels like a great use case for MongoDB and a very awkward fit for a relational database.
The types of queries you would be making against this data is very key to best schema design but imagine that your documents (in a single collection similar to 1. above) look something like this:
{ "round" : 1,
"eventType": "et1",
"attributeName": "attributeValue",
...
}
You can easily query by round, by eventType, getting back all attributes or just a specified subset, etc.
You don't have to know up front how many attributes you might have, which ones belong with which event types, or even how many event types you have. As you build your prototype/application you will be able to evolve your model as needed.
There is a very large active community of Rails/MongoDB folks and there's a good chance that you can find a lot of developers you can ask questions and a lot of code you can look at as examples.
I would encourage you to try it out, and see if it feels like a good fit. I was going to add some links to help you get started but there are too many of them to choose from!
Since you might have a question about whether to use an object mapper or not so here's a good answer to that.
A good write-up of dealing with dynamic attributes with Ruby and MongoDB is here.

How do I avoid complex joins in star schema?

My fact table holds a user score in a course he took. Some of the details of the course, which I have to show on the report, comes from more then one table (in the actual OLTP db).
Do I create a none normalized version of that course entry in a dimension table?
Or do I just join the fact table directly to the course table join to the other tables that describe this course (course_type,faculty who created this course etc)
Snowflaking or bridge tables do make the joins more complicated, and not just from a coding perspective, it also makes it less simple for BI users.
In most cases, I would put these directly in existing or additional dimension tables.
For instance, you have a scores fact table, which has the user details in a dimension which may or may not hold demographics on the user (perhaps it's only a bridge). Sometimes it is better to split out demographic information. So even though the gender and age might be associated with a user entity, in the dimensional model, these might be individual dimensions or lumped into a single dimension - all depending on the usage scenarios.
Perhaps your scores are attached to a state and states have regions (snowflake). It might be far more efficient for analysis to have the region dimension linked directly instead of going through the state dimension.
I think what you will find is that the dimensional model is a very pragmatic denormalization approach. The main things which are non-negotiable are the facts - after that the choice of dimensions is very much informed by the behavior of the data, your foresight for common usage scenarios - and avoiding falling into the too few dimensions and too many dimensions problems.
Maybe I do not understand your question, but a fact table in a star schema is supposed to be joined to dimension tables surrounding it.
If you do not feel like making joins, simply create a view, and use the view for reporting.
If you were to post a model (schema), it would be easier to comment/help.
It is a common practice to consolidate several dimensions together, sacrificing normalization in favor of performance. This is usually done when your typical query will need all dimensions together (as opposed to using different bits for different use cases).
Also remember that while you receive a reduction in join overhead, there are some drawbacks:
Loss of flexibility, which might hinder development as the warehouse expands
Full table scans take longer (in traditional row-based RDBMS such as SQL Server)
Disk space consumption
You will have to consider each case separately.
It might be worthwhile to also consider the option of creating a materialized view, if such ability is offered by your RDBMS.
We commonly have a snowflake schema as the physical DWH design, but add a reporting view layer that flattens the snowflake schema into a star schema.
This way your OLAP cube becomes much simpler adn easier to manage.

Resources