[zeromq-dev] RPC implementation details

[Lucas Hope]

Hi there,

I have used JSON-RPC - one command per message. Multi-part messages are
allowed as a set of commands to be treated as an atomic operation, so if one
fails/errors, the whole batch fails/rolls back. The RPC layer is a thin~ish
API to a database backend.

The architecture is simple REQ-REP clients with an LRU queue connecting
REQ-REP workers. As it is a database app, the database is the bottleneck and
the workers are all local, so I didn't trouble with a more complicated
heartbeating mechanism.

Peace,

-Luke

On Fri, Jul 22, 2011 at 7:32 AM, mrevilgnome <mrevilgnome at gmail.com> wrote:

> Howdy,
>
> I'm in the middle of doing some exploring/prototyping and I was
> curious how others implemented RPC on top of 0mq.  Initially I was
> going to use 0mq as a transport for Avro or Protobuf, but then I
> realized that by doing so I would sacrifice a fair bit of 0mq
> functionality.  My current thinking is that I should use multipart
> messages; a header which is the function name and a value which is the
> protobuf serialized parameters for the call.  Initially I thought that
> I might want two headers one of which would be used for routing
> between clusters by function, but I figure that I can have nodes
> register capabilities with a router.  Does this seem reasonable?  What
> have others done?  Please poke holes if you see them.  Thanks.
>
> --Matt
> _______________________________________________
> zeromq-dev mailing list
> zeromq-dev at lists.zeromq.org
> http://lists.zeromq.org/mailman/listinfo/zeromq-dev
>



-- 
---------------------------------------------------
Dr Lucas Hope - lucas.r.hope at skype
Machine Learning and Software Engineering Consultant
Melbourne, Australia
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20110722/e02d4c01/attachment.htm>