Multicast问题 - 未名空间MITBBS历史存档

国际科技财经博客移民网络热点娱乐民生时事公众号

Redian新闻

>未名空间

>EmergingNetworking - 热门网络技术

Multicast问题

Multicast问题# EmergingNetworking - 热门网络技术

p*a2007-07-10 07:07

1 楼

当multicast forwarding path已经从shared tree switch到source tree之后，原来
receiver到rp之间的(S,G) entry会被prune掉，此时如果receiver到rp之间没有其他
receiver，则默认三分钟后rp上(*,G)对应于相应接口的entry也会被remove掉。
问题:如果此时相对于receiver的rp后端有一个新的该group的source出现，则receiver
无法知晓该新的resource？进而也无法接收该新resource所发出的组播数据？

s*g2007-07-10 07:07

2 楼

Assuming we are talking about IGMPv1/v2 with PIM SM ...
Where did you get the idea that (*,G) will be pruned when traffic switches
from shared tree to source tree? (*,G), as its name implies, includes all
sources for group G, (*,G) will be pruned on shared tree routers ONLY when
last receiver from last hop router leaves the group, which is not the case
in your question.
So back to your question, when new source comes in, traffic will flow
through shared tree from RP to receiver normally, which h

【在 p***a 的大作中提到】

: 当multicast forwarding path已经从shared tree switch到source tree之后，原来
: receiver到rp之间的(S,G) entry会被prune掉，此时如果receiver到rp之间没有其他
: receiver，则默认三分钟后rp上(*,G)对应于相应接口的entry也会被remove掉。
: 问题:如果此时相对于receiver的rp后端有一个新的该group的source出现，则receiver
: 无法知晓该新的resource？进而也无法接收该新resource所发出的组播数据？

p*a2007-07-10 07:07

3 楼

是我的问题描述有误，我起初认为(*,G)条目下相应的OIL会被prune掉
因为当没有child(S，G)情况下默认3min之后该OIL就会timeout，进而被prune掉
事实如你所说，该OIL不会被prune掉，因为receiver keeps sending join message to
rp

Receiver

【在 s*****g 的大作中提到】

: Assuming we are talking about IGMPv1/v2 with PIM SM ...
: Where did you get the idea that (*,G) will be pruned when traffic switches
: from shared tree to source tree? (*,G), as its name implies, includes all
: sources for group G, (*,G) will be pruned on shared tree routers ONLY when
: last receiver from last hop router leaves the group, which is not the case
: in your question.
: So back to your question, when new source comes in, traffic will flow
: through shared tree from RP to receiver normally, which h

t*r2007-07-10 07:07

4 楼

不错的面试题, zher给M一个吧~

z*r2007-07-10 07:07

5 楼

啥时候的帖子？怎么没有印象

【在 t*******r 的大作中提到】

: 不错的面试题, zher给M一个吧~

d*i2007-07-10 07:07

6 楼

can you please see if my understanding is right on this?
on the RP, (*, G) won't be pruned due to periodically (*, G) join sent by
the leaf router if there is receiver in the group, but the specific (S, G)
will be pruned after the switch over to SPT. Is this correct?

Receiver

【在 s*****g 的大作中提到】

z*r2007-07-10 07:07

7 楼

2 scenarios:
1. the RP is not on the shortest path. the upstream router will use a
different interface as the RPF check, and send a prune message to the RP
removing itself from the source tree. so on RP, the interface facing this
upstream router will be removed from the OIL. So your understanding is right.
2. the RP is on the shortest path, (s, g) will not be pruned after switching
over to SPT. of coz this depends on the spt-threshold, if the threshold is
set to infinity, the router will not eve

【在 d****i 的大作中提到】

: can you please see if my understanding is right on this?
: on the RP, (*, G) won't be pruned due to periodically (*, G) join sent by
: the leaf router if there is receiver in the group, but the specific (S, G)
: will be pruned after the switch over to SPT. Is this correct?
:
: Receiver