Wednesday, 15 May 2013

messaging - Why use AMQP/ZeroMQ/RabbitMQ -


Contrary to writing your library.

We are working on a project that is a self-split server pool, if a section becomes very heavy, the manager will divide it and make it a separate machine Keeps in the form. It will also warn all connected clients who affect it to connect to the new server.

I'm curious about using ZeroMQ for inter-server and inter-process communication. My partner would like to roll herself, I am looking at the community to answer this question.

I am a quite novice programmer and have just learned about message lines. As I googled and read, it seems that everyone is using message lines for all things, but why? Does he make it better than writing his library? Why are they so common and why are so many?

What makes them better by writing your own library?

When there is probably nothing to carry out the first version of your app: Your needs are well defined and you will develop a messaging system that meets your needs: Small feature list, small source code etc.

After the first release, very useful later in those devices, when you can actually increase your application and add more features to it Are there. I'll give you some usage cases:

  • Your app needs to talk to a big Andian machine (X86, Intel / AMD) with a large Andian Machine (Spark / PowerPC). Your messaging system had some endurance ordering assumptions: go and fix it
  • You have created your app, so this is not a binary protocol / messaging system, and now it is very slow because you spend most of your time Pars (number of messages have increased and parsing has become an obstacle): Customize it so that it can transport binary / fixed encoding.
  • Start Switch to have had 3 machine within a LAN, any everything from delayed machine. Your client / boss / point-hair-devil-boss appears and tells you that you will install the app on the van which you do not manage - and then you start connection failure, poor latency etc. You need to archive the message and need to resend them later: Go back to the code and plug in this content (and enjoy)

  • were sent The message needs to be answered, but not all of them: you expect a spreadsheet as a result instead of sending some parameters and just sending and accepting, go back to the code and plug in this content (and Enjoy.)

  • Some messages are important and send required requirements as per reception / backup / backup / persistence. Why do you ask

And many other usage cases that I forgot ...

You can apply it yourself, but do not spend too much time doing this: Will replace it later too.


No comments:

Post a Comment