Hi all,

I am using storm for the purpose of parsing real-time messages. I have
found out the workload hasn't been divided smoothly on the cluster. For
example, if I am using 3 workers for a topology, 2 of them have a capacity
of 0.9 and processing almost 100 tuples per second, but the other one has a
capacity of 0.05 and processing almost 5 tuples per second. All workers are
running on a homogeneous Storm cluster, so all the resources are the same.

Is there any way that I can manage the balancing mechanism of workload? How
does the balancing work in a default way?

Regards,
Ali

Reply via email to