Nice observations. It would be interesting to know, if you are running a clustered index environment, whether a CSV lookup, that can be distributed to the indexers would improve the run time if the lookup runs on the distributed indexers rather than all data running on the SH where the KV store is. You are right that the 'where' clause as part of inputlookup is often overlooked, but I have also found it does make a difference, because it filters the data going into the pipeline.
... View more