Spanner stores its actual data in bigtable, so any file based optimizations are unlikely