Hi Diego and Thomas,
Firstly, I want to go through the combination of {Tx, Rx, Shared} and {Hard,
Soft}. I think Tx or Rx can go with both Hard and Soft, but Shared can only go
with Hard. The reason is Shared cell could be used by more than a pair of
nodes, so monitoring function (i.e. self monitoring cell quality and
re-allocate accordingly ) of soft cell will not work. Do you agree?
Secondly, regarding to the cell used for best-effort track, I think both
dedicated Tx/Rx and Shared cell are fine. But, specially, the cells reserved by
OTF for best-effort track should be dedicated Tx/Rx cells. The reason is that
OTF basically focuses on a pair of nodes and wants to get some guaranteed
bandwidth. Make sense?
ThanksQin
On Monday, May 11, 2015 10:32 AM, Prof. Diego Dujovne
<[email protected]> wrote:
Thomas,
From my point of view, regarding the dedicated/shared discussion,
I vote for dedicated TX and RX.
I think the main goal of OTF is to dynamically manage bandwidth allocation
for data transmission purposes; then it is essential to guarantee that this
bandwidth is available with no contention. If contention is enabled (as in
shared cells), bandwidth access becomes statistic, then neither the bandwidth
availability nor the delay can be guaranteed.
If OTF can also be used for management or non-priority traffic, that is, what
I should call, real best-effort traffic, then shared cells can be an option. As
far as I can remember, we have not considered specifically this type of traffic
in our draft discussions to date.
So, I see three options:
-narrow the type of traffic to that requiring only dedicated cells,
-keep OTF traffic-agnostic,
-allow the upper layers (an application, for example) to send a
dedicated/shared flag
together with each bandwidth request. I think that a bandwidth request
accounting
service may be required in this case.
Any thoughts?
Diego
2015-05-10 22:54 GMT-03:00 Thomas Watteyne <[email protected]>:
Diego,I agree with you that OTF is orthogonal to the flags, but I also believe
we should be as precise as possible. Let's forget for a second in which draft
what goes and think only about the end result. What flags would you put on the
OTF cells, and would they be dedicated/shared? I have some weak opinion, but
I'd love to hear your pros/cons.Thomas
On Sunday, May 10, 2015, Prof. Diego Dujovne <[email protected]> wrote:
Thomas,
The only unknown here are the attributes supported
by the best-effort track. Since we are using best-effort, and
from the references I have cited, I think it must be specified
on the 6top draft whether dedicated (TX and RX) cells can
be used on the best-effort track. We have also replaced the
track references to bundle references, so it is clear OTF
is a L3 mechanism, so we specify now "best-effort bundle"
instead of "best-effort track".
AFAIK, we say nothing on the draft about the
cell attributes (shared, TX or RX), and we shall keep saying
nothing, since I think track and bundle attributes are outside
OTF scope.
Regards,
Diego
2015-05-10 1:52 GMT-03:00 Thomas Watteyne <[email protected]>:
Diego,
I dont Fully understand what you are proposing, hence some clarifying questions.
Currently, the OTF algorithm triggers when a node should schedule one cells TO
its neighbor. Since all nodes run the OTF algorithm, the current node can also
receive requests to schedule cells FROM any of its neighbors.
Each cell is associated (this is not OTF specific):- one or more flags: TX, RX,
SHARED- a neighbor, optional- a type: hard or soft
What does the draft say now about the qualifiers, and what are you proposing to
change those to?
Thomas
On Saturday, May 9, 2015, Prof. Diego Dujovne <[email protected]> wrote:
All,
Although we discussed briefly about this issue,
we did not arrive to a conclusion on this topic:
"And I have not found requiring cells with specific link option (i.e., Tx only
or Rx only).Is there any plan to add this feature into OTF?IMHO, I think it may
be useful to take advantage of TSCH that can make contention free links between
nodes.(From Jongsoo Jeong)"
I do agree that with OTF we are working with bandwidth requests for data
traffic,
so dedicated TX or RX links should be preferred. I have not found any specific
restrictions on the link options available for the Best Effort track which is
the
one we are using for dynamic allocation. It is only mentioned on the 6top
sublayer draft as follows:
If the TrackID is set to (0,0), the cell can be used by the best-
effort QoS configuration or as a Shared cell. If the TrackID is not
set to (0,0), i.e., the cell belongs to a specific track, the cell
MUST not be set as Shared cell.
But it does not say that a best effort track MUST use Shared cells.
We also discussed tracks for OTF in June 2014, but there were no comments on
dedicated cells:
http://www.ietf.org/mail-archive/web/6tisch/current/msg02335.html
Any thoughts?
Thanks,
Diego
--
DIEGO DUJOVNE
Académico Escuela de Ingeniería en Informática y Telecomunicaciones
Facultad de Ingeniería UDP
www.ingenieria.udp.cl
(56 2) 676 8125
--
DIEGO DUJOVNE
Académico Escuela de Ingeniería en Informática y Telecomunicaciones
Facultad de Ingeniería UDP
www.ingenieria.udp.cl
(56 2) 676 8125
--
DIEGO DUJOVNE
Académico Escuela de Ingeniería en Informática y Telecomunicaciones
Facultad de Ingeniería UDP
www.ingenieria.udp.cl
(56 2) 676 8125
_______________________________________________
6tisch mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6tisch
_______________________________________________
6tisch mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6tisch
