Hi all, I read from old post of nathan replying to similar question --- > We are trying to migrate one of our realtime workflow from a customized application from a single machine to storm 0.72 with 2 machines(16core, 48G memory). > When we do the testing, we were using the default storm configuration, which means 2 supervisor on 2 machines with 4 workers per machine. > > We are using 1 ScribeSpout to accept logs from our serving system, several bolts to do the log parsing and transformation, another several bolts to do the aggregation. > > When we turn off the acker(by changing the acker number to 0), we found the "Complete latency(ms)" is 0 by storm ui. > When we turn on the acker(using 1 or 2 acker), we found the "Complete latency" is about 700-800ms by storm ui. > > Does this number means the time between getting the log in the spout and all acks received form all the bolts?
Its the time between spout emitting the tuple and the tuple tree being completed. --- yes I get it, is it the time of last tuple or average of all tuples or average of last some tuples? Sincerly, Adrian SJ Lee
