Ok folks, I'm back. I've spent the last six months reading, writing (and arithmetic - that's a joke... ha-ha...) seriously, implementing solutions, visiting customer sites, and seeing what's going on in the world; and to that end, the DW/BI world is changing. You've seen me write about this before, but now it's on-top of your head.
Like it or not... Your Data Warehouse IS a System Of Record, and if it isn't in 2009, it WILL BE before the end of 2010. So what does that mean?
You've heard about it for the past 2 years (at least), it's called Operational BI, and my good friend Claudia Imhoff, and my other friend Colin White have published many many articles and blog entries about these topics. But have you given any consideration to exactly what that means to the Data Warehouse?
Come on people, get with the times!
Ok, down to brass tacks. To have "operational BI" you must have Operational Data, well, wait a minute - operational data means system of record doesn't it?
Ding Ding Ding, you win the prize... come on down!
Wow, didn't know there was a prize... (there's not)... Anyhow, Operational Data is required - and if you're talking operational BI, then your talking about business intelligence, and if you mention business intelligence, then you MUST consider data warehousing....
So what does that mean?
It means you've GOT NO CHOICE but to consider an Operational EDW under the covers. That's right folks, I've said it for years, now I'm not going to hide this fact from the light anymore. YOUR DATA WAREHOUSE WILL FROM NOW ON... HOUSE OPERATIONAL DATA RIGHT BESIDE YOUR HISTORICAL DATA, this means that the operational data will be captured from applications sitting directly ON-TOP of your Data Warehouse.
Think I'm foolin do ya?
Think again... These are the systems I've been building, or helping folks build for the past 3 years - you don't think that I've been sitting around resting on my laurels eh? Far from it. I've been traveling the world, talking to folks about accountability, credibility, RAW OPERATIONAL DATA sitting right inside the data warehouse. This means that the combined result produces Operational BI. it also means that the resulting Data Warehouse or (ODW for short) IS a system-of-record.
Don't like it? Tough. Take it to the bank!
Customers like JP-Morgan-Chase, Lockheed Martin, Raytheon, World Bank, ABN-AMRO, RBS (royal bank of Scotland), Sisters of Mercy Healthcare, Blue-Cross Blue Shield, Department of Defense, Food and Drug Administration, Federal Aviation Administration, City of Charloette (North-Carolina), SNS Bank Netherlands, and more have all used these principles to create complete ODW's in house today.
What do they need?
Accountable, Auditable, Flexible and scalable data stores that act as an EDW while at the same time absorbing real-time transactional (operational) data directly in to the EDW. Thus creating a RAW ODW. It's here today folks, fed by web-services (SOA), batch, and ESB's - managed and edited by applications that control metadata, carrying changes and audit trails along with cube-write-backs and data edits, that's what ODW and Operational BI are all about. It's making the systems more robust, more scalable, more complete.
At the risk of upsetting some folks: And ODW is the underpinning of successful OBI.
The ODW contains and inherits all the responsibilities of the data warehouse. In addition, because it houses operational data it also inherits all the responsibilities of an operational system!!! Yep, you guessed it: 9x9 up-time, reliability, fault-tolerance, fail-over, garaunteed access, high speed transaction handling, on-top of: strategic data handling, historical record keeping, integration, etc... And don't forget: system-of-record.
But why is it a system-of-record? I though other operational systems handled this?
Each operational system handles their own system of record, but with an ODW - think about it... you're saying: My data warehouse/BI solution is an operational system in addition to being a data warehouse. So, you're saying: this is the only place where a complete, single picture of the integrated operational and historical data lives. Which in essence means: you have built what needs to be a system-of-record.
What's involved in creating a system-of-record for the ODW?
Good question, and I wish I had time to answer it.... Ok - here's my opinion: you need to worry about all the operational "rules and regulations and governance", you also need to worry about storing RAW (uncleansed, unchanged, unaltered) data sets in your data warehouse. That means, your "business rules" to mainpulate data can no longer be executed up-stream of your data warehouse. You must move your business rules downstream of the ODW, otherwise you're changing operational data to make it fit, and blowing up any chance of an auditable system-of-record. ESPECIALLY when 3 months go by and the business rules change!!
Are you saying I need "RAW" data in my ODW to make this work?
That's exactly what I'm saying. I'm not saying get rid of the business rules, I'm not saying put "bad" data in front of business users.... No, I am saying "move the processing downstream" to POST-ODW (between ODW and Data Mart delivery). It will make your life MUCH easier.
This is absured, I've never heard of anything like this before. Why should I care?
You don't have to care if you don't want, but the world is a changing place. If you want an auditable and compliant ODW, or you want to add operational data to your data warehouse you will have to follow these rules, dare I say it: you might find yourself out of a job if you don't!
All the companies I mentioned above, and more, every day are learning this the hard way (through failed audits, through corporate restructuring, through spread-marts where cost has gotten out of hand)... They have decided to build ODW's to solve these problems.
It just so happens, they are using the Data Vault Modeling to make this happen, you don't need a Data Vault model to get here, you must however follow the principles outlined in this article - but the Data Vault is an enabler that will help you get here faster. You can read more about the Data Vault Model at: http://www.DataVaultInstitute.com
or contact me directly: DanL@DataVaultInstitute.com
Posted October 28, 2009 12:46 AM
Permalink | No Comments |