13 Sep
2014
13 Sep
'14
10:48 a.m.
Hi.
As some of you already know, the last days I have been testing the
possibility of use 802.11s as link layer instead of adhoc.
By default 11s comes with a kind of layer2 routing (same as bat-adv
does) by using HWMP (hybrid wireless mesh protocol) which in general is
quite shitty compared to other well known solutions.
However, if the option mesh_fwding is set to 0, this l2 routing layer is
disabled and 11s can be used just as layer1 protocol in the same level
of Ad-Hoc or AP/Infrastructure.
From OpenWRt these are the needed options to enable 11s as link layer:
uci set wireless.mesh0.mode=mesh
uci set wireless.mesh0.mesh_id=wibed
uci set wireless.mesh0.mesh_fwding=0
Using 11s instead of adhoc for deploying our mesh networks can bring
some interesting features, among others:
1. Better support for 11n
2. Better compatibility with drivers (probably even ath_htc works fine)
3. You can bridge it to another interface if necessary.
4. It does NOT try to synchronize the TSF counter of your wifi card thus
you can create up to 8 11s VAP mixed with adhoc, AP, client, etc...
5. I don't know deeply 11s but it probably has better design for
deploying mesh networks
I've been testing it in a 12 nodes mesh network mixing it with bmx6 and
it worked as a charm. No more "strange" problems coming from the adhoc
layer.
Yesterday Gio the great added support for 11s in libre-mesh [1]. So now
we only need to test it a bit more with our already crazy network
architecture, so it may became even a bit more crazy for traditional
mesh folks (bmx6+bat-adv+11s, never done something like that!).
Opinion? Concerns? Ideas?
[1] https://github.com/libre-mesh/lime-packages/tree/feature/80211s
--
./p4u
13 Sep
13 Sep
11:59 p.m.
On 9/13/2014 5:48 AM, Pau wrote:
Hi.
Using 11s instead of adhoc for deploying our mesh networks can bring
some interesting features, among others:
1. Better support for 11n
2. Better compatibility with drivers (probably even ath_htc works fine)
3. You can bridge it to another interface if necessary.
4. It does NOT try to synchronize the TSF counter of your wifi card thus
you can create up to 8 11s VAP mixed with adhoc, AP, client, etc...
5. I don't know deeply 11s but it probably has better design for
deploying mesh networks
I like the pros.
What are the cons?
I've been testing it in a 12 nodes mesh network mixing it with bmx6 and
it worked as a charm. No more "strange" problems coming from the adhoc
layer.
Very nice.
Yesterday Gio the great added support for 11s in libre-mesh [1]. So now
we only need to test it a bit more with our already crazy network
architecture, so it may became even a bit more crazy for traditional
mesh folks (bmx6+bat-adv+11s, never done something like that!).
So 11s would be layer1, bat-adv l2, bmx6 l3 ?
14 Sep
14 Sep
11:06 p.m.
I don't know any con yet :) But we should see about the performance of both and
compare.
The layer1 (11s) links you with the direct neighbors, the layer2 is batman-adv but only
for your local cloud or neighborhood, then the layer3 (bmx6) attach all the clouds in a
single big routed network.
On 14/09/14 01:59, Charles N Wyble wrote:
On 9/13/2014 5:48 AM, Pau wrote:
--
./p4u
Hi.
Using 11s instead of adhoc for deploying our mesh networks can bring
some interesting features, among others:
1. Better support for 11n
2. Better compatibility with drivers (probably even ath_htc works fine)
3. You can bridge it to another interface if necessary.
4. It does NOT try to synchronize the TSF counter of your wifi card thus
you can create up to 8 11s VAP mixed with adhoc, AP, client, etc...
5. I don't know deeply 11s but it probably has better design for
deploying mesh networks
I like the pros.
What are the cons?
I've been testing it in a 12 nodes mesh network mixing it with bmx6 and
it worked as a charm. No more "strange" problems coming from the adhoc
layer.
Very nice.
Yesterday Gio the great added support for 11s in libre-mesh [1]. So now
we only need to test it a bit more with our already crazy network
architecture, so it may became even a bit more crazy for traditional
mesh folks (bmx6+bat-adv+11s, never done something like that!).
So 11s would be layer1, bat-adv l2, bmx6 l3 ?
_______________________________________________
Dev mailing list
Dev(a)lists.libre-mesh.org
https://lists.libre-mesh.org/mailman/listinfo/dev
15 Sep
15 Sep
12:09 a.m.
On 14/09/14 20:06, Pau wrote:
I don't know any con yet :) But we should see
about the performance of both and compare.
veeery very promising,
i did a quick test with our 14.08 branch,
between a tl-wdr3500 and a tl-mr3040
and it worked like a charm;
netperf between the two devices 1 meter away from each other
yielded 40mbit :D so 802.11n speeds are working perfectly (contrary to
adhoc mode, which needs to restart "wifi" after boot)
will test further
very nice pau and henning for the tip
cheers!
gui
The layer1 (11s) links you with the direct neighbors, the layer2 is batman-adv but only
for your local cloud or neighborhood, then the layer3 (bmx6) attach all the clouds in a
single big routed network.
On 14/09/14 01:59, Charles N Wyble wrote:
On 9/13/2014 5:48 AM, Pau wrote:
Hi.
Using 11s instead of adhoc for deploying our mesh networks can bring
some interesting features, among others:
1. Better support for 11n
2. Better compatibility with drivers (probably even ath_htc works fine)
3. You can bridge it to another interface if necessary.
4. It does NOT try to synchronize the TSF counter of your wifi card thus
you can create up to 8 11s VAP mixed with adhoc, AP, client, etc...
5. I don't know deeply 11s but it probably has better design for
deploying mesh networks
I like the pros.
What are the cons?
I've been testing it in a 12 nodes mesh network mixing it with bmx6 and
it worked as a charm. No more "strange" problems coming from the adhoc
layer.
Very nice.
Yesterday Gio the great added support for 11s in libre-mesh [1]. So now
we only need to test it a bit more with our already crazy network
architecture, so it may became even a bit more crazy for traditional
mesh folks (bmx6+bat-adv+11s, never done something like that!).
So 11s would be layer1, bat-adv l2, bmx6 l3 ?
_______________________________________________
Dev mailing list
Dev(a)lists.libre-mesh.org
https://lists.libre-mesh.org/mailman/listinfo/dev
5:05 a.m.
On 09/14/2014 09:09 PM, Gui Iribarren wrote:
On 14/09/14 20:06, Pau wrote:
I would very much like to test/compare network performance in both modes
on a live network. It sounds promising indeed.
In particular we have had lots of trouble with highly interconnected
meshes of >50 nodes; I believe this is more of a wi-fi layer problem,
but it would be interesting to see if 11s has a positive impact in those
scenarios.
abrazos,
Nico
I don't know any con yet :) But we should
see about the performance of both and compare.
veeery very promising,
20 Sep
20 Sep
1:27 a.m.
On 14/09/14 21:09, Gui Iribarren wrote:
On 14/09/14 20:06, Pau wrote:
weeeeell... it seems setting mesh_fw=0 is not enough
in this 4 node mesh (miaumiau, alfredo, casapuente and labanda-o)
root@miaumiau:~# iw wlan0-mesh station dump
Station 64:70:02:3d:91:3c (on wlan0-mesh)
inactive time: 0 ms
rx bytes: 736246338
rx packets: 2974686
tx bytes: 205027479
tx packets: 659121
tx retries: 44394
tx failed: 159
signal: -53 [-53] dBm
signal avg: -49 [-49] dBm
Toffset: -57249438374 us
tx bitrate: 65.0 MBit/s MCS 7
rx bitrate: 65.0 MBit/s MCS 7
mesh llid: 1315
mesh plid: 1314
mesh plink: ESTAB
mesh local PS mode: ACTIVE
mesh peer PS mode: ACTIVE
mesh non-peer PS mode: ACTIVE
authorized: yes
authenticated: yes
preamble: long
WMM/WME: yes
MFP: no
TDLS peer: no
Station f8:d1:11:3b:6d:c6 (on wlan0-mesh)
inactive time: 40 ms
rx bytes: 185883890
rx packets: 1772482
tx bytes: 5207025
tx packets: 17410
tx retries: 10163
tx failed: 139
signal: -73 [-73] dBm
signal avg: -74 [-74] dBm
Toffset: -2624074667 us
tx bitrate: 19.5 MBit/s MCS 2
rx bitrate: 26.0 MBit/s MCS 3
mesh llid: 558
mesh plid: 364
mesh plink: ESTAB
mesh local PS mode: ACTIVE
mesh peer PS mode: ACTIVE
mesh non-peer PS mode: ACTIVE
authorized: yes
authenticated: yes
preamble: long
WMM/WME: yes
MFP: no
TDLS peer: no
Station 90:f6:52:c6:00:f0 (on wlan0-mesh)
inactive time: 50 ms
rx bytes: 292257513
rx packets: 2378740
tx bytes: 76005329
tx packets: 153870
tx retries: 37616
tx failed: 116
signal: -49 [-49] dBm
signal avg: -51 [-51] dBm
Toffset: -2540802411 us
tx bitrate: 65.0 MBit/s MCS 7
rx bitrate: 58.5 MBit/s MCS 6
mesh llid: 1113
mesh plid: 1827
mesh plink: ESTAB
mesh local PS mode: ACTIVE
mesh peer PS mode: ACTIVE
mesh non-peer PS mode: ACTIVE
authorized: yes
authenticated: yes
preamble: long
WMM/WME: yes
MFP: no
TDLS peer: no
miaumiau, casapuente and alfredo see a "phantom" batman gateway (labanda-o)
while labanda-o is clearly gw_mode=client, and at best "relaying"
gateway announcement from another node farther in the batman cloud
(64:70:02:4e:c5:80)
so it seems 11s is mangling in some way the mac addresses in packets,
breaking batman-adv running on top of that layer1
root@miaumiau:~# batctl gwl
Gateway (#/255) Nexthop [outgoingIF]: advertised
uplink bandwidth ... [B.A.T.M.A.N. adv 2014.2.0, MainIF/MAC:
eth0-167/02:e6:fc:be:28:9d (bat0)]
64:70:02:4e:c5:80 (192) 64:70:02:3d:91:3c [wlan0_adhoc-167]: 10.0/2.0
MBit
=> 64:70:02:3d:91:3c (239) 64:70:02:3d:91:3c [wlan0_adhoc-167]: 10.0/2.0
MBit
palmeras_eth0-167 (168) 90:f6:52:c6:00:f0 [wlan0_adhoc-167]: 10.0/2.0
MBit
too bad, i'm dismantling the test setup...
anyway, from informal tests running netperf before and after, i didn't
see a significant difference in throughput or stability, between adhoc
and 11s mode
so, it looked very promising but maybe it's not quite "there" yet
the good news is that ath9k adhoc mode is looking astonishingly stable
so far, been running for two weeks with no hangups... Should we send
Felix some chocolates? or maybe cut his hands to keep the code intact?
cheers!
gui
I don't know any con yet :) But we should
see about the performance of both and compare.
veeery very promising,
i did a quick test with our 14.08 branch,
between a tl-wdr3500 and a tl-mr3040
and it worked like a charm;
netperf between the two devices 1 meter away from each other
yielded 40mbit :D so 802.11n speeds are working perfectly (contrary to
adhoc mode, which needs to restart "wifi" after boot)
will test further very nice pau and henning for the tip
cheers!
gui
The layer1 (11s) links you with the direct neighbors, the layer2 is batman-adv but only
for your local cloud or neighborhood, then the layer3 (bmx6) attach all the clouds in a
single big routed network.
On 14/09/14 01:59, Charles N Wyble wrote:
_______________________________________________
Dev mailing list
Dev(a)lists.libre-mesh.org
https://lists.libre-mesh.org/mailman/listinfo/dev
On 9/13/2014 5:48 AM, Pau wrote:
Hi.
Using 11s instead of adhoc for deploying our mesh networks can bring
some interesting features, among others:
1. Better support for 11n
2. Better compatibility with drivers (probably even ath_htc works fine)
3. You can bridge it to another interface if necessary.
4. It does NOT try to synchronize the TSF counter of your wifi card thus
you can create up to 8 11s VAP mixed with adhoc, AP, client, etc...
5. I don't know deeply 11s but it probably has better design for
deploying mesh networks
I like the pros.
What are the cons?
I've been testing it in a 12 nodes mesh network mixing it with bmx6 and
it worked as a charm. No more "strange" problems coming from the adhoc
layer.
Very nice.
Yesterday Gio the great added support for 11s in libre-mesh [1]. So now
we only need to test it a bit more with our already crazy network
architecture, so it may became even a bit more crazy for traditional
mesh folks (bmx6+bat-adv+11s, never done something like that!).
So 11s would be layer1, bat-adv l2, bmx6 l3 ?
_______________________________________________
Dev mailing list
Dev(a)lists.libre-mesh.org
https://lists.libre-mesh.org/mailman/listinfo/dev
1:59 a.m.
On 19/09/14 22:27, Gui Iribarren wrote:
On 14/09/14 21:09, Gui Iribarren wrote:
meh... i rebooted the 4 nodes at once and things went back to normal (?)
will give it one more chance, leaving it running for a few days more.
(rebooting only labanda-o (the "phantom" batman gateway) didn't fix
things)
cheers
On 14/09/14 20:06, Pau wrote:
miaumiau, casapuente and
alfredo see a "phantom" batman gateway (labanda-o)
while labanda-o is clearly gw_mode=client, and at best "relaying"
gateway announcement from another node farther in the batman cloud
(64:70:02:4e:c5:80) I don't know any con yet :) But we should
see about the performance of both and compare.
so it seems 11s is mangling in some way the mac addresses in packets,
breaking batman-adv running on top of that layer1
root@miaumiau:~# batctl gwl
Gateway (#/255) Nexthop [outgoingIF]: advertised
uplink bandwidth ... [B.A.T.M.A.N. adv 2014.2.0, MainIF/MAC:
eth0-167/02:e6:fc:be:28:9d (bat0)]
64:70:02:4e:c5:80 (192) 64:70:02:3d:91:3c [wlan0_adhoc-167]: 10.0/2.0
MBit
=> 64:70:02:3d:91:3c (239) 64:70:02:3d:91:3c [wlan0_adhoc-167]: 10.0/2.0
MBit
palmeras_eth0-167 (168) 90:f6:52:c6:00:f0 [wlan0_adhoc-167]: 10.0/2.0
MBit
too bad, i'm dismantling the test setup...
anyway, from informal tests running netperf before and after, i didn't
see a significant difference in throughput or stability, between adhoc
and 11s mode
so, it looked very promising but maybe it's not quite "there" yet
the good news is that ath9k adhoc mode is looking astonishingly stable
so far, been running for two weeks with no hangups... Should we send
Felix some chocolates? or maybe cut his hands to keep the code intact?
cheers!
gui
very nice pau and henning for the tip
cheers!
gui
_______________________________________________
Dev mailing list
Dev(a)lists.libre-mesh.org
https://lists.libre-mesh.org/mailman/listinfo/dev
The layer1 (11s) links you with the direct neighbors, the layer2 is batman-adv but only
for your local cloud or neighborhood, then the layer3 (bmx6) attach all the clouds in a
single big routed network.
On 14/09/14 01:59, Charles N Wyble wrote:
_______________________________________________
Dev mailing list
Dev(a)lists.libre-mesh.org
https://lists.libre-mesh.org/mailman/listinfo/dev
On 9/13/2014 5:48 AM, Pau wrote:
> Hi.
>
>
>
> Using 11s instead of adhoc for deploying our mesh networks can bring
> some interesting features, among others:
>
> 1. Better support for 11n
> 2. Better compatibility with drivers (probably even ath_htc works fine)
> 3. You can bridge it to another interface if necessary.
> 4. It does NOT try to synchronize the TSF counter of your wifi card thus
> you can create up to 8 11s VAP mixed with adhoc, AP, client, etc...
> 5. I don't know deeply 11s but it probably has better design for
> deploying mesh networks
I like the pros.
What are the cons?
>
> I've been testing it in a 12 nodes mesh network mixing it with bmx6 and
> it worked as a charm. No more "strange" problems coming from the adhoc
> layer.
Very nice.
>
> Yesterday Gio the great added support for 11s in libre-mesh [1]. So now
> we only need to test it a bit more with our already crazy network
> architecture, so it may became even a bit more crazy for traditional
> mesh folks (bmx6+bat-adv+11s, never done something like that!).
So 11s would be layer1, bat-adv l2, bmx6 l3 ?
_______________________________________________
Dev mailing list
Dev(a)lists.libre-mesh.org
https://lists.libre-mesh.org/mailman/listinfo/dev
5:05 p.m.
On 09/20/2014 03:27 AM, Gui Iribarren wrote:
weeeeell... it seems setting mesh_fw=0 is not enough
so it seems 11s is mangling in some way the mac addresses in packets,
breaking batman-adv running on top of that layer1
I was wondering (wrote it in an italian ML, then Ilario asked me to
report it here) how can batman-adv coexist with 802.11s with HWMP disabled.
In this document [1] they describe 802.11s and how it works. A new frame
format with at least 4 addresses is introduced. Now, since we want to
disable HWMP in order to use batman-adv, who's going to fill in those
addresses?
anyway, from informal tests running netperf before and after, i didn't
see a significant difference in throughput or stability, between adhoc
and 11s mode
so, it looked very promising but maybe it's not quite "there" yet
There are several enhancements regarding QoS and multi-channel
operation... I don't know about the rest, maybe it gives the best with
other scenarios like mesh where the distance varies greatly...
[1] http://www.ieee802.org/802_tutorials/06-November/802.11s_Tutorial_r5.pdf
30 Sep
30 Sep
4:44 p.m.
http://lists.open80211s.org/pipermail/devel/2014-September/003462.html
On 20/09/14 19:05, David wrote:
On 09/20/2014 03:27 AM, Gui Iribarren wrote:
--
./p4u
weeeeell... it seems setting mesh_fw=0 is not
enough
so it seems 11s is mangling in some way the mac addresses in packets,
breaking batman-adv running on top of that layer1
I was wondering (wrote it in an italian ML, then Ilario asked me to
report it here) how can batman-adv coexist with 802.11s with HWMP disabled.
In this document [1] they describe 802.11s and how it works. A new frame
format with at least 4 addresses is introduced. Now, since we want to
disable HWMP in order to use batman-adv, who's going to fill in those
addresses?
anyway, from informal tests running netperf before and after, i didn't
see a significant difference in throughput or stability, between adhoc
and 11s mode
so, it looked very promising but maybe it's not quite "there" yet
There are several enhancements regarding QoS and multi-channel
operation... I don't know about the rest, maybe it gives the best with
other scenarios like mesh where the distance varies greatly...
[1] http://www.ieee802.org/802_tutorials/06-November/802.11s_Tutorial_r5.pdf
_______________________________________________
Dev mailing list
Dev(a)lists.libre-mesh.org
https://lists.libre-mesh.org/mailman/listinfo/dev
3508
days inactive
3525
days old
8 comments
5 participants
participants (5)
-
Charles N Wyble -
David -
Gui Iribarren -
Nicolás Echániz -
Pau