sql - Metadata reporting database - what type of DB schema should I use? -

I projected a project for the purpose of collecting server configuration metadata from Windows Server and to store it in DB. I will collect data for more than 100 configuration fields for each server.

In the client who wants to be able to work, it is to compare data configured for the same server at different points of the same time, or two different servers which are the same function (i.e. Exchange Server ) To see if there is a difference and what the difference might be.

For DB design, I usually just normalize all the data in the OLTP type schema, where all the same configuration items will remain on a table related to their specific area (such as hardware information) but I am thinking that this may be a bad move and I should save it to an OLAP type of data warehouse.

I'm not sure how to go with DB design, so it can do with some direction on whether I want to normalize data and create many tables, or with a huge table Should go with normalization and more than 100 fields, or should I look at a star topology or something completely different (EAV)?

I limited to using NET and MSSQL server 2005.

EDIT: The tool will be run on an essential basis instead of grabing config data every day / week to collect and store data. Wants to maintain data for at least some years.

The best for star schema reporting is to use Star Schema for storing the purpose of my experience Not because it can be a set of scenes (indexed for display) and you can later design ideas for star schema, to record the configuration changes, the storage model should have a set of event tables. You can start with the flat log file structure and it can generally be considered normal to search for good storage for storage and queries. The storage model is considered good if you can define the deficiency of the model, then the reporting model should be good for faster ad hoc questions. You should pay attention to the storage model because the reporting model is a denormalization of the storage model and it is easy to denormalize later. EAV structures are useless for both models because you can not define any barrier, but questions are complicated complex.