s**u 发帖数: 17 | 1 给224.0.0.5和224.0.0.6?
谢谢 | c*a 发帖数: 806 | 2 did a quick google search on igmp snooping and 224.0.0.5, and come up with
RFC4541,
and
http://support.microsoft.com/kb/223136 (also referenced in RFC4541)
【在 s**u 的大作中提到】 : 给224.0.0.5和224.0.0.6? : 谢谢
| c******e 发帖数: 82 | 3 OSPF use multicast to exchange link state info among routers. So all OSPF
routers need to join 224.0.0.5 & 224.0.0.6 multicast groups in order to
receive routing updates from others.
【在 s**u 的大作中提到】 : 给224.0.0.5和224.0.0.6? : 谢谢
| b******e 发帖数: 66 | 4 224.0.0.5 is for allospf router, 224.0.0.6 is used for DR router in multi-
access network only. Although they are multicast groups, an OSPF router does
not send out IGMP joins for the groups, when OSPF is enabled on a router's
interface, this interface will be programed to send/receive packets destined
to those IP addresses. Since all happens across one directly connected link
, there is no need to build multicast tree, igmp join/query and pim process
are unnecessary.
【在 c******e 的大作中提到】 : OSPF use multicast to exchange link state info among routers. So all OSPF : routers need to join 224.0.0.5 & 224.0.0.6 multicast groups in order to : receive routing updates from others.
| b******e 发帖数: 66 | 5 I don't understand what RRAS does and not sure why it will ever sends IGMP
joins to 224.0.0.5 or 224.0.0.6
In a pure switch/router network, since ospf router never sends igmp joins (
or rip, hsrp for that matter), igmp snooping on switch does not come into
play anyways, so where is the problem? one thing for sure, there is no such
word as "igmp" ever mentioned in rfc2328.
【在 c*a 的大作中提到】 : did a quick google search on igmp snooping and 224.0.0.5, and come up with : RFC4541, : and : http://support.microsoft.com/kb/223136 (also referenced in RFC4541)
| z**r 发帖数: 17771 | 6 igmp join/report is necessary here, I agree it is not in the original ospf
rfc, however, along with the igmp snooping more and more popular (in most
mid to high end switch, igmp snooping is on by default), if one ospf node
never sends out igmp join/report, this particular ospf node will never get a
chance to receive other ospf information
such
【在 b******e 的大作中提到】 : I don't understand what RRAS does and not sure why it will ever sends IGMP : joins to 224.0.0.5 or 224.0.0.6 : In a pure switch/router network, since ospf router never sends igmp joins ( : or rip, hsrp for that matter), igmp snooping on switch does not come into : play anyways, so where is the problem? one thing for sure, there is no such : word as "igmp" ever mentioned in rfc2328.
| z**r 发帖数: 17771 | 7 only true when igmp snooping is enabled, otherwise, multicast will be
treated like broadcast, so one doens't have to join any mcat group to
receive ospf updates
【在 c******e 的大作中提到】 : OSPF use multicast to exchange link state info among routers. So all OSPF : routers need to join 224.0.0.5 & 224.0.0.6 multicast groups in order to : receive routing updates from others.
| c*a 发帖数: 806 | 8 Guess it's not enough to just list reference, I should have copy/paste the
key points here.
The reason is IGMP Snooping, but only IGMP Snooping in very AGGRESSIVE mode,
which will also snoop on permanent group (i.e. 224.0.0.0/24). Most of IGMP
snoop apply to groups outside of that range, thus not causing problems for
OSPF and other routing protocols, such as EIGRP, RIPv2 and etc.
RRAS is only mentioned as one of multiple vicitims of this "aggressive" mode
of IGMP snooping from certain vendors.
【在 b******e 的大作中提到】 : I don't understand what RRAS does and not sure why it will ever sends IGMP : joins to 224.0.0.5 or 224.0.0.6 : In a pure switch/router network, since ospf router never sends igmp joins ( : or rip, hsrp for that matter), igmp snooping on switch does not come into : play anyways, so where is the problem? one thing for sure, there is no such : word as "igmp" ever mentioned in rfc2328.
| s**u 发帖数: 17 | 9 So that is the key.
We must know the detail about how the igmp snooping works... what kind of
packet it tracks...
Any guy knows ?
thanks
a
【在 z**r 的大作中提到】 : igmp join/report is necessary here, I agree it is not in the original ospf : rfc, however, along with the igmp snooping more and more popular (in most : mid to high end switch, igmp snooping is on by default), if one ospf node : never sends out igmp join/report, this particular ospf node will never get a : chance to receive other ospf information : : such
| b******e 发帖数: 66 | 10 Doing a simple Google search on "igmp snooping" will give what you want, but
here we go, IGMP snooping is a L2 ethernet switch feature where switch can
intercept IGMP report/query messages exchanged between host and L3 router,
the switch then program it MAC forwarding table such that multicast traffic
is only forwarded to ports which have active receiver, without the feature,
multicast traffic will be broardcasted to all ports in the same VLAN. Of
course, all these have to be done in hardware.
T
【在 s**u 的大作中提到】 : So that is the key. : We must know the detail about how the igmp snooping works... what kind of : packet it tracks... : Any guy knows ? : thanks : : a
| z**r 发帖数: 17771 | 11 igmp report/query doesn't have to be between hosts and L3 routers, a L2
switch can also send out igmp snooping querier and expect igmp join/report.
it was my understanding too that in the old time, ospf doesn't need to join
any mcast group. However, the fact is, on most high end or mid range switch,
igmp snooping is on by default for every VLAN, and not every switch vendor
seperates 224.0.0.x groups and treats them like broadcast. With this being
said, OSPF needs to do something in order to take
【在 b******e 的大作中提到】 : Doing a simple Google search on "igmp snooping" will give what you want, but : here we go, IGMP snooping is a L2 ethernet switch feature where switch can : intercept IGMP report/query messages exchanged between host and L3 router, : the switch then program it MAC forwarding table such that multicast traffic : is only forwarded to ports which have active receiver, without the feature, : multicast traffic will be broardcasted to all ports in the same VLAN. Of : course, all these have to be done in hardware. : T
|
|