| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 1297.1 | GIGAswitch is probably using DEC spanning tree algorithm | NACAD2::BATTERSBY | | Thu Aug 11 1994 13:24 | 19 |
| I inquired about this with one of our DB900MX firmware developers,
and I was told that the GIGAswitch firmware rev in your two
GIGAswitches may be using the DEC spanning tree algorithm instead
of the IEEE 802.1d spanning tree algorithm. Apparently the forwarding
delay is computed differently between the two. The DB900MX's are using
802.1d of course. So what happens is that devices implementing 802.1d
spanning tree see certain parameters like forwarding delay coming
from a device (which in this case is a GIGAswitch using DEC spanning
tree value computations), and halves the forwarding delay value
propagating from the GIGAswitch which is the root.
Total Forwarding delay is made up of listening time and learning time.
DEC spanning tree uses this total time (whatever assigned or the
default of 30 seconds), as the value propagated. In IEEE 802.1d, the
Forwarding Delay value propagated is half the value that would be
propagated from a root device implementing DEC spanning tree.
I guess this is one of the subtle differences between 802.1d and DEC
spanning tree implementations.
Bob
|
| 1297.2 | | NPSS::MDLYONS | Michael D. Lyons - Young enough and dumb enough | Thu Aug 11 1994 15:00 | 5 |
| The GIGAswitch has only supported 802.1d from day 1. I'll try to
reproduce this and see what it's sending in the bpdus versus what SNMP
thinks is set.
MDL
|
| 1297.3 | More information | YUPPY::GAY | David Gay - (7)847 6543 | Fri Aug 12 1994 10:56 | 11 |
| All of my DECbridge 900s are set to use the 802.1d spanning tree
algolrithm.
I've changed the root bridge to be a DECbridge 900 and all of the
bridges now report the correct value (7 secs) for the forward delay.
The GIGAswitch, however, now reports a Forward Delay of 14 secs.
Regards,
David.
|
| 1297.4 | Forwarding delay calculation is not being calculated right... | NACAD2::BATTERSBY | | Fri Aug 12 1994 11:23 | 5 |
| it sounds like the managemment interface code within the GIGAswitch
is not using the right formula to calculate and report to the
user the value of forwarding delay.
Bob
|
| 1297.5 | | NACAD::ANIL | | Fri Aug 12 1994 14:24 | 20 |
| The answer is: although the Gigaswitch implements IEEE 802.1D STP,
it interprets forward-delay like we used to in the DEC STP.
A port typically transitions through the following when it first
comes up:
Listen state -> Learn state -> Forwarding state
Time spent in the Listen state (bridge is listening to hellos to
construct a spanning tree) is equal to the amount of time spent
in Learn state (bridge is learning about stations on extended LAN)
which is equal to what the 802.1D spec refers to as forward-delay,
and requires a bridge to put in its BPDU.
As you can see, the total amount of time taken by a bridge port before
it can go into Forwarding is 2*forward-delay. However, Gigaswitch
(and some earlier Digital bridges) defined the total time as forward-delay.
This is why it is off by a factor of 2 from 802's definition.
Anil
|
| 1297.6 | Thanks.... | YUPPY::GAY | David Gay - (7)847 6543 | Mon Aug 15 1994 09:05 | 12 |
| Anil,
Thanks for the explanation.
I'll set the Forward Delay to 14 secs on the GIGAswitch to achieve what
I wanted.
Will this implementation be changed to conform to the standard ?
Regards,
David.
|
| 1297.7 | | NACAD2::ANIL | | Mon Aug 15 1994 18:49 | 3 |
| GIGAswitch engineering will have to answer that one..
Anil
|
| 1297.8 | | NPSS::MDLYONS | Michael D. Lyons - Young enough and dumb enough | Tue Aug 16 1994 10:00 | 1 |
| It's a bug, it'll be fixed.
|