The discussion below is to overview some of the tools in each are we are using or evaluating for future use both internally and for clients.
- Get the Data In – ETL or ELT
ETL Project X has had some great success with Ascential. We have even developed a proprietary development methodology called Xcellerate Data(TM) that helps get the data in faster.
ELT, we see some great promise and possible speed to data opportunties here and as we see all the EDW vendors moving more into the central version of the truth (Oracle and DB2 moving marts onto the warehouse and Teradata is there) this makes even more sense. We are looking to Sunopsis for this.
- Metadata – a real issue and an area that needs some help. Project X does not see anyone filling this need properly and have held out based on upcoming releases from Erwin and Ascential – wait and see.
- The Data Model – the ugly stepchild of enterprise information management. If done right it is a complete model of your business so the business should be able to find the data. So look at buying an industry model from your warehouse provider.
- The Warehouse – this is a religious war and that is not polite around the dinner table, but please aim for the single version of the truth it makes things so much more interesting. But seriously, this is your aircraft carrier which holds major power in your battle for performance and value. Be careful what you buy, you just might get what you think you paid for at twice the expected cost with half the value.
- The View / Aggregates (and even marts)
In Teradata this is how you virtualize the access and aggregation of data in EII like IPEDO and AquaLogic you do the same thing. This is a serious tool in your arsenal as if you create views on the unstructured data you can get the data into the hands of the user, while underneath the covers you can normalize the data with minimal impact to the end user. And if you are fedarated then you will need an EII tool to see all the data around the organization.
- Service enabling the data
This is the tricky part. It is being bantered around and I am coming on side slowly that you should be publishing your views as a services. This does a bunch of things, it gives freedom to those with rights to leverage the data and services are built to see services. It also means that if underlying data structures change then it will minimize impact to the services it supports. But most important, you can register the service and you get the governance value that somes with SOA. Right now I am only seeing TAP from Teradata and the EII vendors playing, so check your functions and features.
- So now we need a Registry for our data service – centralize is the ECM mantra so Systinet comes to the rescue.