Column encoding and compression type in manual projection design

May_1

In a performance tuning effort, we tried to redesign the projections. We started with DBD, but the design that DBD came up with didn't actually improve the query performance, compared to our current design. So we had to start the projection design manually. Are there any tools/functions we can use to determine the best encoding or compression types for all the columns if we know what columns will be in the SORT BY clause of the projection?

[Deleted User]

For compression, I'm not aware of any currently-released tools for this outside of the DBD (though someone may have written one); I'd suggest writing a script that just tries a bunch of compressions on a model of your data and sees which gives the smallest data size. That said, if you can do significantly better than the DBD by hand, we'd very much like to hear about your situation. The DBD is in fact intended to produce an optimal design; if you can do significantly better, that likely means that there is a bug in the DBD. In which case we'd like to fix it. Thanks, Adam