pfreixes · October 15, 2015 20:21
diff --git a/16K amqp messages per second b/16K amqp messages per second
 1 publisher > 1 queue
 queues = 10K
 messages per queue = 100
 concurrence consumers = 20
 queues binded by consumer = 50
 qos by consumer = 40
 asyncronous pattern

 Throughput got 16K messages, why ?

 + Threads are to slow, time(context switch) > time(select)
 + As many consumers more memory conumed but more faster
  + Decrease the number of consumers but increase the QOS => more messages in the socket buffer ready to be picked up
  + Rabbitmq gets randomly the next messages considering multiple binding
    + ^^ this is necessary, otherwise there is again resource contention.
      + this is because we use 1 queue per publisher
	1 publisher > 1 queue
	queues = 10K
	messages per queue = 100
	concurrence consumers = 20
	queues binded by consumer = 50
	qos by consumer = 40
	asyncronous pattern

	Throughput got 16K messages, why ?

	+ Threads are to slow, time(context switch) > time(select)
	+ As many consumers more memory conumed but more faster
	+ Decrease the number of consumers but increase the QOS => more messages in the socket buffer ready to be picked up
	+ Rabbitmq gets randomly the next messages considering multiple binding
	+ ^^ this is necessary, otherwise there is again resource contention.
	+ this is because we use 1 queue per publisher