Hi team, I have a general query. Given schema, tot...
# general
s
Hi team, I have a general query. Given schema, total no of rows in Pinot table and some cardinality information per column, I wanted to see how much will be approximate Pinot table size on disk.
Because its a columnar format, I was thinking
Copy code
if dictionary enabled:
    add dict size + (cardinality * max length/bytes)
if not
   take actual bytes for storage(eg 4 for int/ 8 for long etc) * no of rows
Is there something I am missing?
s
Copy code
add dict size + (cardinality * max length/bytes)
This is not correct byte size of fwd index bit encoded = (numbits * numDocs) / 8 + byte size of dictionary = cardinality * length (fixed or var)
plus you need to add inv index etc etc
m
Yes, as Sid mentioned, it is bit encoded. But you can get that from cardinality too.
s
thank you!