[zeromq-dev] Does ZMQ "Over Send" Using OpenPGM

[Steven McCoy]

On 27 October 2010 18:17, Martin Sustrik <sustrik at 250bpm.com> wrote:

> On 10/20/2010 06:31 PM, Steven McCoy wrote:
>
>  Multiple multicast groups are required for scaling up, i.e. reducing the
>> recovery burden for each transport, and to take advantage of network
>> filtering - splitting the application domain across multicast groups
>> such that only a small number are in use by a receiving node at any one
>> time.
>>
>
> Yes. That's how I understood it.
>
> Now, there's a problem with multiple multicast groups (getting multiple
> copies of messages if port numbers collide). I still don't understand what
> it is caused by.
>
> Does it have anything to do with multicast loopback? In other words, if we
> disable multicast loopback, will the problem disappear?
>
> If not so, what's the exact scenario that leads to receiving single packet
> multiple times?


 Incoming packets with the same port, the multicast address is ignored by
the networking stack.

Using separate data-destination port would help but you are still filtering
packets in user space, it is better to push the filtering down the stack for
UDP encapsulation and use UDP's destination port.

-- 
Steve-o
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20101027/e8cd6489/attachment.htm>