On Sun, Jul 25, 2010 at 10:38 AM, Oliver Smith <[email protected]> wrote:

> That's kind of the issue I was pondering, but I actually thought that
> they were more similar. In both cases it's just a snapshot at a certain
> point.

There's no "just"... :-)  In a market data scenario the snapshot is
only for the instruments you've subscribed to.  Thus to send a
snapshot means iterating through the whole topic tree that's been
subscribed to.

> Because my dataset is market-data like, my methodology is just going to
> be to have each client connect first to the PUB/SUB socket, then fetch
> the latest snapshot from the REQ/REP socket, and then begin recv()ing
> from the PUB/SUB socket, so that it may receive old updates on the
> pub/sub connection but it won't miss any that occured between it
> requesting the snapshot and beginning recv()ing the SUB socket.

That'll work.  You are in effect building a distributed snapshot that
is updated asynchronously.  This could be built as a reusable
application, and IMO comes close to what distributed hashmaps do.  I
don't see this as a 0MQ device but rather an application (or a
library).  Could be quite a useful app.

-Pieter
_______________________________________________
zeromq-dev mailing list
[email protected]
http://lists.zeromq.org/mailman/listinfo/zeromq-dev

Reply via email to