[rabbitmq-discuss] one thread is consistently slower than another

[Emile Joubert]

Hi Jonathan,

It is not clear how the broker could be responsible for making client
threads perform better or worse, so this might be a Python threading or
perhaps a Pika issue rather than a broker issue. Do all consumers use
the same message prefetch count?

You say the client threads interact with a networked filesystem and that
they publish. Have you tried to break down which of the client subtasks
are slower? Publishing an asychronous operation, so all threads should
take the same amount of time for that subtask. When you measure publish
speed, what exactly are you measuring?




-Emile