Database Reference
In-Depth Information
Global grouping
Global grouping makes sure that the entire stream from the source component (spout or
bolt) goes to a single instance of target bolt, to be more precise and specific to the instance
of the target bolt with the lowest ID. Well let's understand the concept with an example,
let's say my topology is as follows:
I will assign the following parallelism to the components:
Also, I will use the following stream groupings:
Then, the framework will direct all data from the myboltA stream instances, that are emit-
ting onto one instance of myboltB stream, which would be the one to which Storm has as-
signed a lower ID while instantiation:
As in the preceding figure, in the case of global grouping, all tuples from both instances of
Bolt A would go to Instance 1 of Bolt B , assuming it has a lower ID than Instance 2 of
Bolt B .
Search WWH ::




Custom Search