SQL_Server 2012 - Columnar Indexes or Netezza

We will be moving to SQL Server 2012 shortly and we are kicking off a major BI Project. We will be consolidating to a single source of truth using a dimensional database (a Star with some snowflaking). The planning phase is progressing nicely with different vendors presenting their ideas for best practices and on how to get there. So far the vendors are in the same direction - a staging area in 3NF feeding a series of datamarts. A Datamart is defined as a star (or STAH as they say locally in Boston) will be on the same server.

The "how" is where the vendors disagree. One vendor is a SQL Server purest (these guys are really, really good) and thinks we can it all in SQL Server (but not SSIS - they are purests after all). The other vendor is OK with Informatica although they would prefer Data Stage because they are IMB guys.

The other area of disagreement is Netezza vs just using Columnar Indexes. The IBM guys are pushing Netezza while the SQL Server Dudes are pushing SQL Server.

At this point, I see the 2 vendors as being equal. One is stronger in some areas and one stronger in other areas.

It seems to me that it's an issue of expertise. Michaelangelo can paint a masterpiece with a spray gun while I would have trouble drawing stick figures with Picasso's brushes. But what do you think?

I've been working with Netezza exclusively for the last 4 years, primarily with retailers with very large databases. It is a very easy system to set up and administer and can achieve extremely high query performance with very little tuning. It is also has some of the fastest load times provided you structure your processes to handle data using set operations (i.e. update all rows at once versus one at a time).

Columnar databases can perform very well, but tend to have much longer load times (based on my experience with Sybase IQ). In addition to transaction logging and other typical overhead of traditional databases, the index structures are many and require time to maintain. If you are dealing with large load volumes I would be concerned with such an environment.

I get the sense that with a smaller database, Columnar Indexes will suffice. But with large databases, Netezza is preferable. Of course, the definition of smaller and larger is changing all the time.