[rabbitmq-discuss] Migrating from MSMQ:Identifying queue features

[Emile Joubert]

Hi,

On 02/01/13 16:59, mabra wrote:
> I have several apps A, which generates messages with expiration time
> of about 2 hours. I have several apps B, which evals these Messages.
> They are sometimes running, sometimes not.

It is possible to define per-queue or per-message timeouts:
http://www.rabbitmq.com/ttl.html

> Usually the B type is running permanently and in normal operation, it
> reads each message, but does this by peek-ing the the queue and leave
> them in the queue.

It is possible to dequeue and requeue a message, which is similar to
peeking at the head of the queue. Bear in mind that repeating this
operation in the absence of other consumers will lead to the same
message being observed.

For rejecting using the .net client see
http://www.rabbitmq.com/releases/rabbitmq-dotnet-client/v3.0.1/rabbitmq-dotnet-client-3.0.1-client-htmldoc/html/type-RabbitMQ.Client.IModel.html#method-M:RabbitMQ.Client.IModel.BasicReject%28System.UInt64,System.Boolean%29

Using rejection of messages to implement message selection in consumers
is not recommended. If different consumers (app B) process different
messages then you are advised to route message based on that difference
so that consumers only ever observe messages that they are able to process.

> With MSMQ, each message has a unique id, so, if my app restarts, it
> can just peek tha last message it has been processsed and then
> continue peeking the remaining messages.

Publishers (app A) could add a unique ID to messages in a header when
publishing which can be used in the same way.

> I concurrent to the B app, on demand another B is comming to the
> same queue, evealuating the same messages with our rules.

It is possible for multiple consumers to read from the same queue. See
tutorial 2:
http://www.rabbitmq.com/tutorials/tutorial-two-java.html

> The is probbaly more a CEP app with a fast temporary store [to avoid 
> databases]. This works exceptionally fast with MSMQ and I am abel to
> read more then 15k messages/s.

RabbitMQ is generally able to exceed that frequency, though it depends
greatly on consumers keeping up, the messaging pattern and hardware
being employed.

For notes about performance see these blog posts:
http://www.rabbitmq.com/blog/2012/04/17/rabbitmq-performance-measurements-part-1/
http://www.rabbitmq.com/blog/2012/04/25/rabbitmq-performance-measurements-part-2/

> A good note, hint or ref to a doc or something could really help me a
> lot. I am using C#/mono.

You may find this book of interest:
http://manning.com/videla/

RabbitMQ has ships with a mature C# client:
http://www.rabbitmq.com/releases/rabbitmq-dotnet-client/v3.0.1/rabbitmq-dotnet-client-3.0.1-user-guide.pdf



-Emile