Alexandr, >> Are there any performance pros or cons?
I would do some benchmarking under realistic data load before going forward with this data model. I don't have numbers to back this up, but if the column in question is in the key, I'd be concerned about the cost of comparing and sorting blobs over integers. This could make read and flush operations, among others, unnecessarily expensive. Performance concerns aside: > So here is almost no any difference between them Like Romain, I'd be surprised if the (very small) disk footprint benefit were worth the extra complexity and loss of clarity in client code. Jim Witschey
