[zeromq-dev] "Devices" nodes and single point of failure

[Berganovsky, Michael]

Hi Martin

I think I understand your suggestion, however setting p2p connections strikes me as a bit "un-elegant" since configuration of such pipeline (or rather processing node in the pipeline) may quickly become a nightmare :) Are you planning some "location" or "directory" services for version 2? 

Regards
Michael Berganovsky

-----Original Message-----
From: zeromq-dev-bounces at lists.zeromq.org [mailto:zeromq-dev-bounces at lists.zeromq.org] On Behalf Of Martin Sustrik
Sent: Wednesday, June 23, 2010 8:52 AM
To: 0MQ development list
Subject: Re: [zeromq-dev] "Devices" nodes and single point of failure

Hi Michael,

> I am new to 0MQ so please excuse me if this topic was already covered 
> somewhere. We are trying to understand how current design of 0MQ 
> (version 2 I believe – development branch)  is handling faults (or 
> scaling) of device nodes (like streamer or forwarder). In other words we 
> are trying to build a pipeline and would like it to be resilient to 
> hardware failures. Nodes which have streamers seems to be “single 
> points” of failure in the design. Am I missing something here? A 
> document or a pattern may be? Any pointers would be greatly appreciated.  

You can eliminate streamer devices by connecting each step 1 processing 
component to each step 2 processing component, without having the device 
in the middle.

Step1---Step2
      \ /
       X
      / \
Step1---Step2

Is that what you had in mind?

Martin

_______________________________________________
zeromq-dev mailing list
zeromq-dev at lists.zeromq.org
http://lists.zeromq.org/mailman/listinfo/zeromq-dev