[zeromq-dev] Federated Binary Star Pattern

Richard Smith rsmith at rsbatechnology.co.uk
Tue Aug 30 15:02:03 CEST 2011


I am interested in exploring further the idea of brokerless reliability, in
both request-reply and pub-sub topologies.  I would really appreciate if
someone could maybe clarify / expand on the 2nd sentence in this paragraph
from the ZGuide (in this section:
http://zguide.zeromq.org/page:all#Preventing-Split-Brain-Syndrome) relating
to Split Brains :

<quote>
We must not split a Binary Star architecture into two islands, each with a
set of applications. While this may be a common type of network
architecture, we'd use federation, not high-availability failover, in such
cases.
</quote>

A typical failover use case for my Clients is a two-stage approach: Fast
failover to secondary server in same data centre as primary, and then a
failover to a Disaster Recovery site.  Let's say the requirements to use
Binary Star architecture as described in the Guide could be applied to the
primary->secondary (assume can share same physical switch etc) case, how
does the idea of federation help in the "two islands" use case (e.g between
primary and DR data centres, or between geographically distant servers)?  I
take it that "federation" in this context was meant as simulation of client
by a worker in the other "island"?

Regards,

Richard
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20110830/09a25085/attachment.htm>


More information about the zeromq-dev mailing list