Had an opportunity to test this new feature released into Public Preview just this month! More to come for sure!
I am currently building exactly that approach at my customer. And generally I love that approach.
But I found out that Streams seem to have their problems when they have too many columns.
So a workaround is to split the streams and underlying views in several streams/views with less columns
A few figures here with a Large WH:
select * from STR_V_VBAP_all – >400 cols
–15070 in 8:25 mins
–29040 in 11:01 mins
–43291 in 9:24 mins
–27k in 8:00 mins
–75K in 8:50 mins
select * from STR_V_VBAP_TEST100; – 100 cols
–26 in 1:15 mins
–197 1:32 mins
–8K in 1:41 mins
–70K in 1:48 mins
–96k in 2:33 mins
select * from STR_V_VBAP_TEST50_LNK; – 50 cols
–60 0:36 mins
–144 0:42 mins
–7.1k in 45s
–66K in 44s
–91K im 47s
select * from STR_V_VBAP_TEST0_LNK; – only key cols ( <10)
–68 in 0:12 s
–822 in 0:14 s
–4100 in 20 s
–5,5k in 13 s
–52K in 19 s
–72K in 16s
Streams are offsets on the underlying object — there isn’t a limitation on the number of columns used, in fact that doesn’t even factor in the discussion of deploying or using stream objects