Row-count is not the only statistic you can use in computing your cost
function. You can use many statistics in your cost function (e.g. selectivity,
column cardinality, predicates).
So, if you decide to use rows-per-minute as your cost function, you can compute
rows-per-minute of the join
Hi Julian,
Thank you for you reply.
I think the problem with interpreting rowCount as the rate is we need both
rate and window size of the inputs for the estimation of output rate of the
RelNodes, and this cannot be embedded into a single number.
As an example, let A and B be two windowed
On Fri, Jul 12, 2019 at 5:43 PM Julian Hyde wrote:
> In practice, the rowCount is just a number. So you can think of it as
> rows-per-second if you are optimizing a continuous query.
>
> If you are using a table in a streaming query, does it have a “rows per
> second?”. Yes - it is the number of
In practice, the rowCount is just a number. So you can think of it as
rows-per-second if you are optimizing a continuous query.
If you are using a table in a streaming query, does it have a “rows per
second?”. Yes - it is the number of rows in the table multiplied by the number
of times per
Looking forward for the outcome :)
Below a few comments regarding the extensibility concerns of Kenn.
In order to find the best plan the VolcanoPlanner just needs to know if one
cost is less than another cost [1] and this is encapsulated in the
isLe/isLt methods [2].
Adding a new cost class
Dear Stamatis,
Thank you for your reply. I will probably go with overriding
computeSelfCost() as the first step. I checked it, and it seems to be
working.
Dear Kenn,
The cited paper estimates those two values for each node and passes it up
but they are not the cost. The cost of a node depends on
Following this discussion, I have a question which I think is on topic.
Seems like there's two places that from my brief reading are not quite
extensible enough.
1. RelNode.computeSelfCost returns RelOptCost has particular measures built
in. Would Alireza's proposal require extensibility here to
Hi Alireza,
Cost models for streams is a very cool topic but I don't have much
knowledge in the domain.
Regarding the implementation details if you have custom physical operators
then it makes sense to implement computeSelfCost() function as you see fit.
Another option is to plug in your custom
Dear Members of Calcite Community,
I'm working on Apache Beam SQL and we use Calcite for query optimization.
We represent both tables and streams as a subclass of
AbstractQueryableTable. In calcite implementation of cost model and
statistics, one of the key elements is row count. Also all the