Use or abuse of ODS as source

Hi,

I am just starting out with a DWH project. Also for me it is the first time.

We already have an ODS. Now I have already read that using ODS as a source for DWH is fine. However it is now suggested to use the ODS as the only source. For doing so all systems should make a complete copy of all their data to ODS. In that way all data needed or not needed can be found in the ODS. For filling the DHW database selections have to be made out of ODS.

The advantages I can think of :
Flexability, ease of having one source,
The disadvantages are: Lack of scalability, unnessecary use of resources on network, datastore, backup and processing power.

Like to hear your opionings on this matter.

Thanks

The ODS people like everything to go through the ODS but the DW people could care less. Yes, if there are multiple transaction systems with huge amounts of data that needs to be combined, then sure - the ODS is find. But if it's not transactional data, then why does it need to go through the ODS? If it's marketing type data purchased from a 3rd party or Goals used in reporting, what's the point of running it through the ODS in order to get to the DW?

Jeff Smith wrote:The ODS people like everything to go through the ODS but the DW people could care less. Yes, if there are multiple transaction systems with huge amounts of data that needs to be combined, then sure - the ODS is find. But if it's not transactional data, then why does it need to go through the ODS? If it's marketing type data purchased from a 3rd party or Goals used in reporting, what's the point of running it through the ODS in order to get to the DW?

Hi Jeff,

Thanks for your response.
In this case the DWH people want to put everything thourgh the ODS. Basically what they want to do is to use the ODS as stagening area. As I understand your correctly this is fine. However the DWH people do not just want to put the data needed in the ODS but just everyting and than figure out what data they actually want/need.

Hans

Sounds to me like either someone has their terminology confused, or, a really messed up architecture in mind.

An ODS is a data store used to address focused operational requirements. It is not indented to be a comprehensive enterprise data store. Data in an ODS tends to be relatively current without significant history.

The comprehensive, long term data store is the data warehouse.

Now, an ODS can be leveraged as a staging area for feeds into the data warehouse, but it is done on an opportunistic basis. In other words, you have already done a lot of work extracting and conforming the data to get it into the ODS, why not pass it through to the DW? But it becomes counterproductive if you are moving new data, which is of no interest in the ODS, through the ODS just for the sake of doing so.

What should be considered is leveraging the data warehouse to provide supplemental information to the ODS application.

Yes, if there are multiple transaction systems with huge amounts of data that needs to be combined, then sure - the ODS is find. But if it's not transactional data, then why does it need to go through the ODS? If it's marketing type data purchased from a 3rd party or Goals used in reporting, what's the point of running it through the ODS in order to get to the DW????