| >(1) Can we add the new 900MX to the FDDI ring (by draging the ABport
> of the 900MX to the FDDI) ON-LINE , without impacting their
> running systems ? This is the custome's most concern.
I guess it depends on what protocols/applications you are running over
the ring.
If you had only DECconcentrators in the hub, I would say probably
yes. Because you have DECswitch 900EF, the answer is no.
The FDDI ports on the DECswitch 900EF will reset and go into
pre-forwarding state, during which time no ring traffic will pass
through the switch. This takes 30 seconds by default.
>(2) Is the dual ring broken (both ring) at the time the reconfiguration?
> Would the ring be partitioned during at that time?
While the new concentrator is being inserted between two other
modules, the ring will wrap.
>(3) Any numerical data describing the non-reponse (or ring broken time)
> time of the ring? (or even the hub?)
>
The hub will take 1-3 seconds to break the old connections and
make the new connections necessary to insert the new module.
During this time the ring will wrap.
More important is the 30 seconds pre-forwarding the 900EF will take
due to spanning tree.
-Mike
|
| Mike,
First thanks for your reply. Your answer is great.
I just want to clarify a bit more.
(1) My understanding is, if the dual-ring is composed of ONLY
900MXs in the DEChub 900MS, when I insert a new 900MX
to the ring, the ring will be wrapped for 1-3 seconds.
So during that time , there are two separate single-ring.
After the 3 seconds, the dual-ring will be restored.
If the running protocols in the network are DECnet (IV and OSI)
and TCP/IP and no real-time applications are running, I don't
think this three seconds would result in somethings abnormal.
What is your comments?
(2) For the customer's case (with 900EF),
> The FDDI ports on the DECswitch 900EF will reset and go into
> pre-forwarding state, during which time no ring traffic will pass
> through the switch. This takes 30 seconds by default.
So during the 30 seconds the 900EF rebuild the forwarding databases
for the FDDI ports, is "during which time no ring traffic will pass
through the switch" meaning that during that time ONLY the Ethernet
nodes would be unable to communicate with the FDDI nodes, all the
FDDI nodes in the dual-ring(after the 3 seconds) would communicate
well ?
What about the state of the Ethernet ports during the 30 seconds
and the effects on the communication between any Ethernet nodes?
Thanks once again.
Feynman.
[Posted by WWW Notes gateway]
|
| Hi -
Mike's on a plane to Interop, so I'll try to finish up this thread.
> If the running protocols in the network are DECnet (IV and OSI)
> and TCP/IP and no real-time applications are running, I don't
> think this three seconds would result in somethings abnormal.
> What is your comments?
These protocols should recover without a hitch. Your DECnet node
event logs may show adjacency down/adjacency up.
> So during the 30 seconds the 900EF rebuild the forwarding databases
> for the FDDI ports, is "during which time no ring traffic will pass
> through the switch" meaning that during that time ONLY the Ethernet
> nodes would be unable to communicate with the FDDI nodes, all the
> FDDI nodes in the dual-ring(after the 3 seconds) would communicate
> well ?
The ring itself is intact. Traffic passing from a 900MX through the
900EF to another 900MX still gets through. However, FDDI traffic that
is destined for or coming from one of the 900EF's Ethernet segments will
not get through.
> What about the state of the Ethernet ports during the 30 seconds
> and the effects on the communication between any Ethernet nodes?
Traffic will still pass between any of the 900EF's Ethernet segments
during this 30 second period, and will pass to any Ethernet LANs
you have configured on the hub backplane.
Does your customer have Ethernet-based VMS Clusters (LAVCs)? This
non-routed protocol is time sensitive. I forget exactly how many
seconds it takes until service to the cluster satellites is dropped,
but it is less than a minute. If the cluster host is reachable
only through FDDI from the nodes on the 900EF's Ethernet segments,
the cluster nodes will hang. When connectivity to the load host
is restored, they will get a new image and eventually be back on
line.
Dotsie
|