Thanks for your interest in MapD.
You can’t create indexes in MapD as we focus on fast scan performance (not saying it won’t ever happen, but its not a current focus area). Unfortunately our way of handling large projection queries (i.e. non-grouped queries) has always needed some optimization. The good news is that with the next release, a change will land that will make these sorts of queries much faster (commit here). Then there is another optimization that will hopefully land in the release after that that should make it even faster.
Note that these optimizations won’t speed up the initial time of loading data from disk, but rather make our in-memory projections more efficient. We have some ideas on selective fetching from disk as well, but likely won’t get to that for a bit.