[zeromq-dev] PGM multicast

tradermbox-zmq at yahoo.com tradermbox-zmq at yahoo.com
Tue May 11 13:32:22 CEST 2010


can we clarify one point.

at the moment with ZMQ-2.0.6, can we run udp (such as udp://eth0.... as a connection address) or we must run epgm://eht0... since i dont think that i can using PGM. Furthermore, is it right to say that EPGM is the protocol using OpenPGM software implementation but PGM is more of a native protocol? 

BTW I have tried the new release and do not see much difference. I am still seeing dropped packets (BTW i am testing subscriber and publisher on the same box..)

Victor




________________________________
From: Steven McCoy <steven.mccoy at miru.hk>
To: 0MQ development list <zeromq-dev at lists.zeromq.org>
Sent: Tue, May 11, 2010 4:15:19 AM
Subject: Re: [zeromq-dev] PGM multicast

On 11 May 2010 16:04, Martin Sustrik <sustrik at 250bpm.com> wrote:

>
>tradermbox-zmq at yahoo.com wrote:
>
>
>> would you say that the only stable transports are tcp and inproc? and
>>> epgm is questionable?
>
>Steven should comment on the stability status of OpenPGM.
>
>>However, it should be noted that multicast transports are inherently
>>complex to set up and are often fail due to inadequate networking
>>hardware, incorrect HW/OS setup etc.
>
>

For example my Alteon is dropping any packet with IP Router Alert (RFC 2113) enabled but my HP ProCurve is handling it fine.  Due to this it will now be a configurable option, similar to SmartPGM.  By default it is disabled, as with version 1 and 2.

UDP/PGM is always more stable given non-conducive network infrastructure

There are no known issues with OpenPGM 2.1.26 aside of unsupported PGM extensions: FEC, Polling, Congestion Control, designated local repairs, and back-off interval auto-tuning.

-- 
Steve-o
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20100511/91357568/attachment.htm>


More information about the zeromq-dev mailing list