Re: [RPRWG] Frame reorder in 802.17b
I believe that the use of bi-directional flooding is not
precluded for wrapping rings. Thus, that would be one
possible solution.
> -----Original Message-----
> From: owner-stds-802-17@listserv.ieee.org
> [mailto:owner-stds-802-17@listserv.ieee.org] On Behalf Of
> Leon Bruckman
> Sent: Tuesday, November 16, 2004 6:56 PM
> To: STDS-802-17@listserv.ieee.org
> Subject: [RPRWG] Frame reorder in 802.17b
>
> Assume the following ring: S1-S2-S3-.....-S64
>
> S1 is transmitting a frame with user DA=A, and the user with
> MAC=A is connected to S64. S1 does not have an entry in the
> SAS DB for A, so it floods it (assume unidirectional flood)
> through ringlet0 (S2, S3, ....
> S64).
> Eventually S64 receives the frame and it reaches user A. User
> A responds, and S64 forwards the frame to S1.
>
> - If we optimize the path by using minimum hops as the direct
> path: S64 sends the frame to S1 directly through ringlet0,
> and the next frame from
> S1 to A will use ringlet1 with a direct frame since it has
> already learned that A is represented by S64. Reorder will
> result since old frames from S1 to A may be in flight in the
> long ringlet0 flooded path.
>
> - A trivial solution would be for S1 to continue using
> ringlet0 with direct frames. This is suboptimal since S1 uses
> the longest path.
>
> - Using bidirectional flooding, with the right cleave point
> solves the problem, but in wrapping rings unidirectional
> flood is used, and even for steering rings unidirectional
> flooding is simpler (no duplication).
>
> Comments please.
>
> Leon
>