You may want to consider using R (RHIPE for large datasets - R with Hadoop MapReduce) for analysis and visualization. R/RHIPE can easily handle 5 million obser

You may want to consider using R (RHIPE for large datasets - R with Hadoop MapReduce) for analysis and visualization. R/RHIPE can easily handle 5 million observations. I have used RHIPE with 500GB dataset. I am currently trying to setup a multi-node RHIPE cluster to handle larger datasets for my personal project. You may be able to use Tableau Public, but your dataset may exceed the storage limits on Public free account. Another option may be to use Hadoop-as-a-Service (no visualization) such as Quoble or TreasureData. I recently tried out TreasureData, a decent service if your data has timestamps. You can run HIVE queries to filter and narrow your data and then export out results to use in Visualization softwares.