Hello,
i've got some doubts about IGMP.
Shall a switch forwaring any membership report /leave group to every
port
(exception the port he heard this messages) or shall he send it
only to a port where he formerly received a igmp query message?
Printable View
Hello,
i've got some doubts about IGMP.
Shall a switch forwaring any membership report /leave group to every
port
(exception the port he heard this messages) or shall he send it
only to a port where he formerly received a igmp query message?
On May 30, 11:27 am, NetworkNerd <k.sperl...@teleos-web.de> wrote:[color=blue]
> Hello,
> i've got some doubts about IGMP.
>
> Shall a switch forwaring any membership report /leave group to every
> port
> (exception the port he heard this messages) or shall he send it
> only to a port where he formerly received a igmp query message?[/color]
---------------------------------------
Firstly by the membership query , which is periodicallly send by the
querier router , and this message is reached to each host now then the
host which want to participate in the multicasting group is , send the
member-ship report to router by-a switch now at that time the
multicate enteries are saved in switch table accordingly , and at
last when a host want to leave the muliticast group shall send a leave
report to router and by this the entery of its in the switch table
also get vanished.........
.....
Vikrant
On May 30, 2:27*am, NetworkNerd <k.sperl...@teleos-web.de> wrote:
[color=blue]
> Shall a switch forwaring any membership report /leave group to every
> port
> (exception the port he heard this messages) or shall he send it
> only to a port where he formerly received a igmp query message?[/color]
First, IP multicast addresses are mapped algorithimically into MAC
multicast addresses.
The default way a L2 switch forwards MAC multicasts is to flood them
to all active ports. So any deviation from this flooding of multicasts
at L2 would only occur if the switch is doing something clever, such
as IGMP snooping (RFC 4541) or, less likely,GMRP (IEEE 802.1D Clause
10).
IGMP snooping and GMRP are used by switches to learn which MAC hosts
(switch ports) care about which MAC multicast addresses, so they can
avoid flooding when it makes sense to avoid flooding.
So:
With IGMPv1 and IGMPv2, the IGMP join reports are sent to the IP
multicast address of the multcast group in question. By default, these
will be flooded to all active ports of the switch. This was done
intentially, to minimize the number of membership reports in a normal
network.
But if IGMP snooping is used by the switch, then the switch only sends
the membership reports to ports it knows are used by multicast
routers.
IGMPv2 leave reports are sent to the all routers multicast group. So
again, the default behavior of the switch is to flood, but if the
switch is using IGMP snooping, it sends leave reports only to the
multicast routers. Also, the switch will stop forwarding the
multicasts to that individual host that has sent in the leave report,
after a configurable amount of time.
In IGMPv3, all reports are sent to a special IGMPv3 routers multicast
group. The dafult behavior is for the L2 switch to flood, but the L2
switch will only send these reports to ports on which multicast
routers are attached, if using IGMP snooping.
Bert
On 30 Mai, 22:39, Albert Manfredi <bert22...@hotmail.com> wrote:[color=blue]
> On May 30, 2:27*am, NetworkNerd <k.sperl...@teleos-web.de> wrote:
>[color=green]
> > Shall a switch forwaring any membership report /leave group to every
> > port
> > (exception the port he heard this messages) or shall he send it
> > only to a port where he formerly received a igmp query message?[/color]
>
> First, IP multicast addresses are mapped algorithimically into MAC
> multicast addresses.
>
> The default way a L2 switch forwards MAC multicasts is to flood them
> to all active ports. So any deviation from this flooding of multicasts
> at L2 would only occur if the switch is doing something clever, such
> as IGMP snooping (RFC 4541) or, less likely,GMRP (IEEE 802.1D Clause
> 10).
>
> IGMP snooping and GMRP are used by switches to learn which MAC hosts
> (switch ports) care about which MAC multicast addresses, so they can
> avoid flooding when it makes sense to avoid flooding.
>
> So:
>
> With IGMPv1 and IGMPv2, the IGMP join reports are sent to the IP
> multicast address of the multcast group in question. By default, these
> will be flooded to all active ports of the switch. This was done
> intentially, to minimize the number of membership reports in a normal
> network.
>
> But if IGMP snooping is used by the switch, then the switch only sends
> the membership reports to ports it knows are used by multicast
> routers.
>
> IGMPv2 leave reports are sent to the all routers multicast group. So
> again, the default behavior of the switch is to flood, but if the
> switch is using IGMP snooping, it sends leave reports only to the
> multicast routers. Also, the switch will stop forwarding the
> multicasts to that individual host that has sent in the leave report,
> after a configurable amount of time.
>
> In IGMPv3, all reports are sent to a special IGMPv3 routers multicast
> group. The dafult behavior is for the L2 switch to flood, but the L2
> switch will only send these reports to ports on which multicast
> routers are attached, if using IGMP snooping.
>
> Bert[/color]
Hi Bert, Hi Vicky,
thanks for answering
the facts you told are well known to me. Maybe my question was not
clearly enough
demonstrated.
If a switch with enabled igmp snooping sends a membership report only
to 'router port'
(i.e. to querrier pos.)
then multicast traffic can only flow in a strength hierachical order:
All multicast traffic
feed into network (i.e. multimedia data) must(!) be feeded in on the
switch which is
also the querier.
Otherwise it can't work.
Am i right?
On Jun 2, 1:44 pm, NetworkNerd <k.sperl...@teleos-web.de> wrote:[color=blue]
> On 30 Mai, 22:39, Albert Manfredi <bert22...@hotmail.com> wrote:
>
>
>[color=green]
> > On May 30, 2:27 am, NetworkNerd <k.sperl...@teleos-web.de> wrote:[/color]
>[color=green][color=darkred]
> > > Shall a switch forwaring any membership report /leave group to every
> > > port
> > > (exception the port he heard this messages) or shall he send it
> > > only to a port where he formerly received a igmp query message?[/color][/color]
>[color=green]
> > First, IP multicast addresses are mapped algorithimically into MAC
> > multicast addresses.[/color]
>[color=green]
> > The default way a L2 switch forwards MAC multicasts is to flood them
> > to all active ports. So any deviation from this flooding of multicasts
> > at L2 would only occur if the switch is doing something clever, such
> > as IGMP snooping (RFC 4541) or, less likely,GMRP (IEEE 802.1D Clause
> > 10).[/color]
>[color=green]
> > IGMP snooping and GMRP are used by switches to learn which MAC hosts
> > (switch ports) care about which MAC multicast addresses, so they can
> > avoid flooding when it makes sense to avoid flooding.[/color]
>[color=green]
> > So:[/color]
>[color=green]
> > With IGMPv1 and IGMPv2, the IGMP join reports are sent to the IP
> > multicast address of the multcast group in question. By default, these
> > will be flooded to all active ports of the switch. This was done
> > intentially, to minimize the number of membership reports in a normal
> > network.[/color]
>[color=green]
> > But if IGMP snooping is used by the switch, then the switch only sends
> > the membership reports to ports it knows are used by multicast
> > routers.[/color]
>[color=green]
> > IGMPv2 leave reports are sent to the all routers multicast group. So
> > again, the default behavior of the switch is to flood, but if the
> > switch is using IGMP snooping, it sends leave reports only to the
> > multicast routers. Also, the switch will stop forwarding the
> > multicasts to that individual host that has sent in the leave report,
> > after a configurable amount of time.[/color]
>[color=green]
> > In IGMPv3, all reports are sent to a special IGMPv3 routers multicast
> > group. The dafult behavior is for the L2 switch to flood, but the L2
> > switch will only send these reports to ports on which multicast
> > routers are attached, if using IGMP snooping.[/color]
>[color=green]
> > Bert[/color]
>
> Hi Bert, Hi Vicky,
> thanks for answering
> the facts you told are well known to me. Maybe my question was not
> clearly enough
> demonstrated.
> If a switch with enabled igmp snooping sends a membership report only
> to 'router port'
> (i.e. to querrier pos.)
> then multicast traffic can only flow in a strength hierachical order:
> All multicast traffic
> feed into network (i.e. multimedia data) must(!) be feeded in on the
> switch which is
> also the querier.
> Otherwise it can't work.
> Am i right?[/color]
--------------------------------------
Hi,
switch (l-2) is not be any type of querier.
For a multicast data flow to be start ... it is first to set a
multicast group, and the multicast group
is formed by the IGMP query and membership query.
I tell u what a switch do in the multicast sceniro...
It is just make enteries in its addressing table (ATU address
translation table ) so that it (l 2 switch)
just pass the data either from host to router or vice versa.......
On Jun 2, 4:44*am, NetworkNerd <k.sperl...@teleos-web.de> wrote:
[color=blue]
> the facts you told are well known to me. Maybe my question was not
> clearly enough
> demonstrated.
> If a switch with enabled igmp snooping sends a membership report only
> to 'router port'
> (i.e. to querrier pos.)
> then multicast traffic can only flow in a strength hierachical order:
> All multicast traffic
> feed into network (i.e. multimedia data) must(!) *be feeded in on the
> switch which is
> also the querier.
> Otherwise it can't work.
> Am i right?[/color]
I agree with Vicky.
Normally, the L2 snoopping switches identify the router ports using a
combination of methods, as described in RFC 4541. Having this info,
the snooping switches can send IGMP reports only to the router ports.
As opposed to normal IGMPv1 and v2, where the reports also go to all
hosts that are members of the multicast group.
In NORMAL IGMPv1 and v2, the hosts that see IGMP join reports from
another group member can suppress their own join report, to reduce
background traffic. But this is not good for IGMP snooping, because
the L2 switches could think that all those non-reporting hosts are no
longer group members. Right? So the snooping L2 switch *wants* to hear
reports from every host.
Now that the L2 switch knows exactly which of its ports has host group
members, it can forward the multicast packets (which arrive from any
of the multicast router ports) only to those ports where member hosts
exist.
From your question, I'm not sure if you are confusing the roles of
the multicast routers and the snooping L2 switches? Remember that in a
network with many multicast routers, only one of these will be elected
as the querier. But the multicast packets themselves may arrive from
more than just one router.
Bert
On Jun 2, 4:44*am, NetworkNerd <k.sperl...@teleos-web.de> wrote:
[color=blue]
> If a switch with enabled igmp snooping sends a membership report only
> to 'router port'
> (i.e. to querrier pos.)
> then multicast traffic can only flow in a strength hierachical order:
> All multicast traffic
> feed into network (i.e. multimedia data) must(!) *be feeded in on the
> switch which is
> also the querier.
> Otherwise it can't work.[/color]
Maybe you are asking something different here. Maybe you are asking
whether a multicast packet sent by one oif the group members can also
find its way to all members of the group?
In other words, are you asking whether multicast packets can be
transmitted by the multicast router AND ALSO by the hosts themselves,
since after all, IP multicast is a many-to-many model?
Yes, why not? Once the L2 switches know where the multicast routers
are and where the host members of the group are, what's to prevent the
L2 switch from sending any packet with that IP multicast destination
address to all the correct places?
Bert