Reputation: 185
I want to benchmark my topology with different parallelism_hint. In order to test it, i prepare a 1G log file to emit to Storm. The problem is how can i measure the time taken by my topology programmatically
Upvotes: 0
Views: 104
Reputation: 62835
time like latency or what? You can see latencies for every component on Storm UI.
Also, there is a "capacity" column -- if you have bottlenecks that you need to scale, their capacity will be close or higher that 1.
This statistics is also can be fetched programatically via Storm REST api.
Upvotes: 1
Reputation: 62350
If you know how many tuples your spout is going to emit and if you have fault-tolerance enabled, you can simple count all incoming acks... Basically, take a startup timestamp in Spout.open(...)
, count the call to Spout.ack()
and if the count reaches the input size, take a final timestamp.
Upvotes: 0