qpid-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Adel Boutros <Adelbout...@live.com>
Subject [Dispatch router 0.6.2] Performance regression of the producers throughput
Date Tue, 04 Oct 2016 08:39:16 GMT
Hello,


In our performance tests, we use 4 Producers (3 JMS Producers + 1 Proton-c Producer) which
all send messages synchronously to a messaging system composed of 1 Dispatcher connected to
2 Qpid Java Brokers (6.0.4). Producers send messages to a Topic which is bound on a queue
(Binding Key + JMS selector filter: 67% of the events sent are accepted by the queue). 3 JMS
Consumers are also attached to consume messages. The queue and topic are set on both brokers
and the dispatch router advertises them correctly.


Connection Schema


Producers - - - - - |                                        |--- Broker 1
(3 JMS + 1 C++)   |                                        |
                              |----- Dispatch Router---|
Consumers -  -  - |                                        |--- Broker 2
(3 JMS)

Dispatch Router has 4 connectors to each Broker.
Dispatch Router + Brokers are on the same machines. Consumers + Producer are on the other
machine.

We have noticed a reduced overall average throughput of the producers by 3% by simply updating
the version of the Dispatch Router from 0.6.1 to 0.6.2 (The release candidate).

Although the regression is tolerable but we would like to have your input on the above knowing
that there are only 4 fixes between the above-mentioned releases.

Regards,
Adel

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message