Skip to contentSkip to conversation

batman_v default throughput

DISCUSSION Moritz Warning 3 participants 5 messages Aug 8, 2019 – Aug 11, 2019

lore
Share & tools
lore
#0 lore raw

Hi,

the Freifunk network in Ulm has a batman_V network of >80 nodes. [0]

For the map, we use the link throughput value from batman-adv to calculate a link value (capped at 40 MBit/s for 100%). The default for VPN connections seem to be 1 MBit/s - when batman-adv doesn't know the throughput from the driver.

Is there a way to the set this throughput or can batman-adv measure it?

Thanks!

[0]: https://map11.freifunk-ulm.de/meshviewer

#1 lore raw

Hi Moritz,

Quoted 9 lines · Moritz Warning

Hi,

the Freifunk network in Ulm has a batman_V network of >80 nodes. [0]

For the map, we use the link throughput value from batman-adv to calculate a link value (capped at 40 MBit/s for 100%). The default for VPN connections seem to be 1 MBit/s - when batman-adv doesn't know the throughput from the driver.

Is there a way to the set this throughput or can batman-adv measure it?

I think we discussed this topic in all its details (with Marek) during WMBv10, no?

However, if you need to refresh the concept, you can look at the function that takes care of selecting the throughput:
batadv_v_elp_get_throughput() in bat_v_elp.c

The function is invoked for each neighbour/link across a given interface and can be summarized as follows:
- if any throughput_override was set by the user for this iface, just use this value;
- otherwise, if the iface is wifi, ask the wifi stack for the expected-throughput (computed by minstrel for ath9k and other SoftMAC drivers);
- otherwise, ask the ethtool module if there is a known link speed for this iface;
- otherwise, if none of the above worked, use the default throughput (1Mbps).

As also mentioned at WBMv10, Marek and I had worked on a patchset that enables the tp_meter to measure the speed of a link, when no other throughput is available (i.e. for VPN or other virtual links). The patchset is still pending here [1].

We suggested you to manifest your interest about that patchset, so that the discussion could move on, but it seemed you were not motivated enough (at least back then).

Maybe you could take this chance to revive that thread? :-)

Cheers,

[1]
https://www.mail-archive.com/[email protected]/msg17833.html

Quoted 4 lines

Thanks!

[0]: https://map11.freifunk-ulm.de/meshviewer

Signature · 1 line
Antonio Quartulli
#2 lore raw
Quoted 4 lines · On Friday, 9 August 2019 10:26:24 CEST Antonio Quartulli wrote:

As also mentioned at WBMv10, Marek and I had worked on a patchset that enables the tp_meter to measure the speed of a link, when no other throughput is available (i.e. for VPN or other virtual links). The patchset is still pending here [1].

It isn't pending. All patches of this series [1] are marked as "changes requested" since almost a year.

The only pending patches are the ones you didn't review yet [2].

Kind regards,
Sven

[1] https://patchwork.open-mesh.org/project/b.a.t.m.a.n./list/?series=157&submitter=&state=*&q=&archive=both&delegate=
[2] https://patchwork.open-mesh.org/project/b.a.t.m.a.n./list/?series=&submitter=&state=&q=&archive=both&delegate=7

#3 lore raw

Hi,

Quoted 9 lines · Sven Eckelmann
Quoted 4 lines · On Friday, 9 August 2019 10:26:24 CEST Antonio Quartulli wrote:

As also mentioned at WBMv10, Marek and I had worked on a patchset that enables the tp_meter to measure the speed of a link, when no other throughput is available (i.e. for VPN or other virtual links). The patchset is still pending here [1].

It isn't pending. All patches of this series [1] are marked as "changes requested" since almost a year.

You're right - I used the wrong word.
When I said "pending" I meant that some $work was done, patches are now idling. Hence my call for "a motivational email" on that thread :)

Thanks for the clarification though!

Quoted 2 lines

The only pending patches are the ones you didn't review yet [2].

oh, thanks for the reminder!

Cheers,

Quoted 6 lines

Kind regards,
Sven

[1] https://patchwork.open-mesh.org/project/b.a.t.m.a.n./list/?series=157&submitter=&state=*&q=&archive=both&delegate=
[2] https://patchwork.open-mesh.org/project/b.a.t.m.a.n./list/?series=&submitter=&state=&q=&archive=both&delegate=7

Signature · 1 line
Antonio Quartulli
#4 lore raw
In reply to Antonio Quartulli

Hi Antonio.

Quoted 7 lines · Antonio Quartulli

Hi Moritz,

Quoted 4 lines · Moritz Warning

Hi,

the Freifunk network in Ulm has a batman_V network of >80 nodes. [0]

Quoted 8 lines
Quoted 5 lines

For the map, we use the link throughput value from batman-adv to calculate a link value (capped at 40 MBit/s for 100%). The default for VPN connections seem to be 1 MBit/s - when batman-adv doesn't know the throughput from the driver.

Is there a way to the set this throughput or can batman-adv measure it?

I think we discussed this topic in all its details (with Marek) during WMBv10, no?

Could be. I do not remember what was discussed..

Quoted 19 lines

However, if you need to refresh the concept, you can look at the function that takes care of selecting the throughput:
batadv_v_elp_get_throughput() in bat_v_elp.c

The function is invoked for each neighbour/link across a given interface and can be summarized as follows:
- if any throughput_override was set by the user for this iface, just use this value;
- otherwise, if the iface is wifi, ask the wifi stack for the expected-throughput (computed by minstrel for ath9k and other SoftMAC drivers);
- otherwise, ask the ethtool module if there is a known link speed for this iface;
- otherwise, if none of the above worked, use the default throughput (1Mbps).

Quoted 4 lines

As also mentioned at WBMv10, Marek and I had worked on a patchset that enables the tp_meter to measure the speed of a link, when no other throughput is available (i.e. for VPN or other virtual links). The patchset is still pending here [1].

That might be a nice fallback instead of employing a cranky iperf setup.
But (imho) it should be off/not compiled in by default and maybe only perform the check once, so to not to scare people into thinking batman-adv consumes too much bandwidth for testing.

Quoted 6 lines

We suggested you to manifest your interest about that patchset, so that the discussion could move on, but it seemed you were not motivated enough (at least back then).

Maybe you could take this chance to revive that thread? :-)

Hm, I do not have that email in my email client anymore. But it's good to know that there is a patch set that I can apply.

Quoted 14 lines

Cheers,

[1]
https://www.mail-archive.com/[email protected]/msg17833.html

Quoted 4 lines

Thanks!

[0]: https://map11.freifunk-ulm.de/meshviewer