Re: Column-oriented database on top of Hadoop: How does that work for user?
ORC and Parquet files, popular file formats in the Hadoop world, are columnar. In large part, the hadoop/spark world has adopted columnar. Vertica can use it's own format, ORC or Parquet. You can even do JOINs between the formats with ease.
Deep analysis of LOTS of data is what it's best at doing. However, you can do single look-ups using optimations provided. So, for example, if you're looking for a single user's information in a sea of electric meter data, there are live aggregate projections and other features that optimize that query.