Currently we create bad sink for every partition here. We could improve this by creating only one instance of a bad sink (per worker node) in the singleton object and reuse it while mapping partitions. Similar approach to e.g. iglu resolver singleton.
Currently we create bad sink for every partition here. We could improve this by creating only one instance of a bad sink (per worker node) in the
singletonobject and reuse it while mapping partitions. Similar approach to e.g. iglu resolver singleton.