We're currently evaluating other database engines outside of MSSQL (currently 2008 R2) for both compiling and querying; developers are at the stage where they want to utilize the existing SSIS/ETL environment as much as possible, so any NoSQL or advanced SQL language is off the cards for now.
I've been thinking of Hadoop/Hortonworks for a while now and I just need a little proof to justify the next step - so, what i was considering was a single node installation of hortonworks against a pre-existing SQL server, both using the same server model (PE 2950). This gives it an opportunity to have a reasonable comparison on both compiling and querying. I'm aware that using SQOOP to run SSIS isn't allowing the full benefit of the hadoop system, but I believe I still get the advantages of Hive - which would include columnstore, which is more...