[Rstplib-users] RE: RSTP Ring Size ?

[Liwei Y. <Liw...@wo...>]

Les,

You did explain things well, but I gauss that you didn't take any =
possible link outage into account.
For example, the 45 bridges ring is only theoretically possible.  =
Because if in worst case, the root bridge lost one of its links, then =
the total age will much more than 23, it will never converge since the =
bpdu aged out at middle of the broken ring !!!

So as a workable ring, the maximum bridges allowed is about 21, unless =
you either assume a perfect ring (why do you need the spanning tree for =
then?) or change the standard on bpdu age calculation. =20

Liwei Yuan

-----Original Message-----
From: Les Bell [mailto:Les...@eu...]
Sent: Monday, February 04, 2002 7:16 AM
To: Robert Petroff
Cc: std...@ie...; rst...@li...
Subject: Re: RSTP Ring Size ?






The 'problem' is caused by the way effective age of a BPDU is =
incremented at
each Bridge, as it is received on its Root Port and subsequently =
propagated on
its Designated Ports - see 802.1w 17.19.19 and 17.19.21.

"IThe effective age of the port information (portPriority and portTimes) =
is
taken as the value of the Message Age parameter carried in a received =
BPDU,
incremented by the greater of (1/16th Max Age) and 1 s, and rounded to =
the
nearest whole second."

This results in the effective age being incremented by 1 at each hop, =
for MaxAge
<=3D 23; an increment of 2 at each hop for 24 <=3D MaxAge <=3D 39; and =
an increment of
3 at each hop for MaxAge =3D=3D 40.

The network size is limited when the effective age of the received info =
from a
BPDU exceeds MaxAge.  The received info must also be refreshed before it =
is aged
and discarded, so the maximum network size is also affected by the value =
of
HelloTime.  The limit of the network is reached when
((effectiveAge+HelloTime)>=3DMaxAge).

With default values of MaxAge (20) and HelloTime (2), the maximum =
network size
18 Bridge hops from the Root Bridge, giving a total of 37 Bridges in a =
chain or
a ring, if the Root Bridge is at the centre.  Setting (MaxAge=3D24) =
causes an
increment of 2 in root times at each hop, so the limit is reached after =
only 11
hops from the Root Bridge, with either (HelloTime=3D1) or =
(HelloTime=3D2).  For
(MaxAge=3D39) and (HelloTime=3D1), the limit is 19 hops from the Root =
Bridge.

The maximum achievable network size is with (MaxAge=3D23) and =
(HelloTime=3D1), with
22 hops from the Root Bridge giving a total of 45 Bridges in a chain or =
a ring.
In practice, you will probably have to reduce this to 21 hops from the =
Root
Bridge (43 in a chain or a ring) to allow for timer variances between =
the
Bridges, as you do not want the Bridges furthest from the Root Bridge to =
age out
the information from its Root Port if the next BPDU is a few =
milliseconds late.

Les...





Robert Petroff <rob...@ya...> on 04/02/2002 11:58:05

Sent by:  Robert Petroff <rob...@ya...>


To:   std...@ie...
cc:   rst...@li... (Les Bell/GB/3Com)
Subject:  Re: RSTP Ring Size ?





Would anyone like to ask on this (see below) my
question, please ?
I realy need it...
Robert

 --- Robert Petroff <rob...@ya...> wrote:
>
> Hi,
>
> We hoped, that IEEE 802.1w (RSTP) will be able
> to solve the problem of a big ring of Bridges.
>
> (In the our legacy STP (IEEE 802.1d) bridges we had
> to "break" the spec.: we allow to exceed the value
> 40
> for MaxAge).
>
> From the first view, we saw, that a new protocol is
> not so sensitive to the LAN diameter, because the
> message age behaves much more like a hop count
> in RSTP than it did in STP.
>
> But now, it seems to us, the state is worse :(
>
> I read the messages in your mailing list:
> a) of Richa Malhotra on
> http://www.ieee802.org/1/private/email/msg00221.html
> with subject "802.1w -- Message/Max age"
> b) of Alex Ruzin on
> http://www.ieee802.org/1/private/email/msg00423.html
> with subject "802.1w: How does LAN diameter depend
> on MaxAge ?"
>
> It seems, that in IEEE 802.1w (RSTP) even our trick
> with MaxAge will not help to increase LAN diameter.
>
> May anyone here provide with clarification, please ?
>
> I humbly thank you in advance, Robert
>
>
>