netone-isolateion branch (NetJack1 for JACK2) first tests

[Stéphane Letz]

Hi,

I've tested netone-isolateion (NetJack1 for JACK2) here in 2 different  
configurations on a 100 MB network (shared with other usages) :

- 2 Linux machines

- One OSX, one Linux

It does not work well here : doing a simple loopback  (sending a sine  
signal generated with "jack_simple_client" from the master to the  
slave and back on the master to hear the result) I hear regular  
"crack" in the sound (like every second...). Using the defauts  
settting, changing the master buffer size, changing the latency (-l)  
parameter does not help.

I don't see any xrun message on any kind of error message on either  
side of the connection.

(For Geoff who said it worked for him, when doing the same loopback  
with an Ardour session, the sound distorsion is much harder to detect,  
so I suggest to use a the simple "sine" test for now...)

Using NetJack2 on the same network works without problems, so I don't  
think the problem is a bad network or whatever.

Now I'm waiting for Jörn and Torben to go on into this "documentation/
understanding/cleaning the code"  work before going further.

Stéphane
_______________________________________________
Jack-Devel mailing list
Jack-Devel@...
http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org

[Jörn Nettingsmeier-5]

Stéphane Letz wrote:
> Now I'm waiting for Jörn and Torben to go on into this
> "documentation/understanding/cleaning the code"  work before going further.

well, netjack1 has been demonstrated to work very well, so i guess the
issues you are seeing are integration problems...

netjack2's qualities notwithstanding, i think netjack1 is the way to go,
and that's what i'll be spending my effort on (particularly because it
has the potential to scale from low-bandwidth WAN via CELT to local LAN
at high channel counts).



--
Jörn Nettingsmeier

Meister für Veranstaltungstechnik

Audio and event engineer
Ambisonic surround recordings

http://stackingdwarves.net
[phone]

_______________________________________________
Jack-Devel mailing list
Jack-Devel@...
http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org

[Stéphane Letz]

Le 28 oct. 2009 à 21:42, torbenh a écrit :

> On Wed, Oct 28, 2009 at 09:56:04AM +0100, Stéphane Letz wrote:
>> Hi,
>>
>> I've tested netone-isolateion (NetJack1 for JACK2) here in 2
>> different configurations on a 100 MB network (shared with other
>> usages) :
>>
>> - 2 Linux machines
>
> does it happen with linux too ?

Yes, it happens also between the 2 Linux machines.

>
>>
>> - One OSX, one Linux
>
> hmm.. oh well, maybe there is still a bug in the packetloss deadline.
> but maybe the packets really come late ?
>
> puzzeling what you actually did.
>

Could you possibly add some debug code to trace what happens?

Thanks

Stéphane
_______________________________________________
Jack-Devel mailing list
Jack-Devel@...
http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org

[torbenh]

On Wed, Oct 28, 2009 at 09:51:44PM +0100, Stéphane Letz wrote:

>
> Le 28 oct. 2009 à 21:42, torbenh a écrit :
>
> >On Wed, Oct 28, 2009 at 09:56:04AM +0100, Stéphane Letz wrote:
> >>Hi,
> >>
> >>I've tested netone-isolateion (NetJack1 for JACK2) here in 2
> >>different configurations on a 100 MB network (shared with other
> >>usages) :
> >>
> >>- 2 Linux machines
> >
> >does it happen with linux too ?
>
> Yes, it happens also between the 2 Linux machines.

hmm... well it seems that it was working ok with -l1
with -l1 i only got this artifact, when the master jackd xruns.

but the deadline calibration was broken, it considered some packets
lost, and went on, with last periods data.

xruns on slave only showed up with -v
now it only happens when the master jackd produces an xrun.

deadline now seems to be floating fine now.

this kind of stuff might still happen on osx, because it doesnt have
ppoll, and maybe i need to round up in the wait function.

anyways, i report slave netxruns with jack_error now. so they should
show up.

>
> >
> >>
> >>- One OSX, one Linux
> >
> >hmm.. oh well, maybe there is still a bug in the packetloss deadline.
> >but maybe the packets really come late ?
> >
> >puzzeling what you actually did.
> >
>
> Could you possibly add some debug code to trace what happens?
>
> Thanks
>
> Stéphane

--
torben Hohn
http://galan.sourceforge.net -- The graphical Audio language
_______________________________________________
Jack-Devel mailing list
Jack-Devel@...
http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org

[Stéphane Letz]

Le 29 oct. 2009 à 01:11, torbenh a écrit :

> On Wed, Oct 28, 2009 at 09:51:44PM +0100, Stéphane Letz wrote:
>>
>> Le 28 oct. 2009 à 21:42, torbenh a écrit :
>>
>>> On Wed, Oct 28, 2009 at 09:56:04AM +0100, Stéphane Letz wrote:
>>>> Hi,
>>>>
>>>> I've tested netone-isolateion (NetJack1 for JACK2) here in 2
>>>> different configurations on a 100 MB network (shared with other
>>>> usages) :
>>>>
>>>> - 2 Linux machines
>>>
>>> does it happen with linux too ?
>>
>> Yes, it happens also between the 2 Linux machines.
>
> hmm... well it seems that it was working ok with -l1
> with -l1 i only got this artifact, when the master jackd xruns.
>
> but the deadline calibration was broken, it considered some packets
> lost, and went on, with last periods data.
>
> xruns on slave only showed up with -v
> now it only happens when the master jackd produces an xrun.
>
> deadline now seems to be floating fine now.
>
> this kind of stuff might still happen on osx, because it doesnt have
> ppoll, and maybe i need to round up in the wait function.
>
> anyways, i report slave netxruns with jack_error now. so they should
> show up.
>
>

Ok, works OK on 2 Linux machines with default parameter for both sides  
of the connections (like "jack_netsbource -h IP and jackd -R -d  
netone"), but :

- I see *3* audio ports on both sides ?? why is that? would be 2 a  
better default value?

- changing the port number to 4 or 8 does not works correctly (like  
"jack_netsbource -h IP -P4 -C4" and  "jackd -R -d netone -i 4 -o4"). I  
get lot of netxruns on both sides and sound is crappy. (same when  
changing the latency parameter)

(BTW: why having different naming for *same* parameters on both  
side ?? (that is  -P -C for in/out channels on master side and -i -o  
on slave). It would help very much to have same naming as much as  
possible...)

Stephane
_______________________________________________
Jack-Devel mailing list
Jack-Devel@...
http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org

[Jörn Nettingsmeier-5]

Stéphane Letz wrote:

>
> Le 29 oct. 2009 à 01:11, torbenh a écrit :
>
>> this kind of stuff might still happen on osx, because it doesnt have
>> ppoll, and maybe i need to round up in the wait function.
>>
>> anyways, i report slave netxruns with jack_error now. so they should
>> show up.
>>
>>
>
>
> Ok, works OK on 2 Linux machines with default parameter for both sides
> of the connections (like "jack_netsbource -h IP and jackd -R -d
> netone"), but :
>
> - I see *3* audio ports on both sides ?? why is that? would be 2 a
> better default value?

? 2 *is* the default value.

what you are seeing is two audio ports and one default midi port.

> - changing the port number to 4 or 8 does not works correctly (like
> "jack_netsbource -h IP -P4 -C4" and  "jackd -R -d netone -i 4 -o4"). I
> get lot of netxruns on both sides and sound is crappy. (same when
> changing the latency parameter)

then something is amiss with your setup... are both your endpoints
low-latency systems? of course in the presence of xruns, the sound is
"crappy" - that's hardly surprising, is it? thus, no points for
constructive criticism.

the thing is, you shouldn't be getting xruns, especially not at those
ridiculously low channel counts.

i've been running a 64ch session at 64 frames latency for hours
yesterday, and it xrunned only when a) there were significant network
traffic peaks unrelated to netjack, b) one of the endpoints would switch
cpu frequency, or c) the master would have a hardware xrun.
it's a best effort thing, no guarantees. but under controlled
conditions, there were no xruns.

> (BTW: why having different naming for *same* parameters on both side ??
> (that is  -P -C for in/out channels on master side and -i -o on slave).
> It would help very much to have same naming as much as possible...)

yeah, that is something i've been nagging torben about, too. patch is
underway. but that's cosmetics imho.

let's have netone in the main repo asap, and then we can work on the
details. my approach would be to eliminate slave configuration
completely and have jack_netsource dictate all parameters to the slave.

iiuc, the current situation arose because torben has been experimenting
with runtime reconfiguration of some parameters, and this means the
slave needs some fixed settings of its own... the final configuration
approach will evolve over time.


best,

jörn








--
Jörn Nettingsmeier

Meister für Veranstaltungstechnik

Audio and event engineer
Ambisonic surround recordings

http://stackingdwarves.net
[phone]

_______________________________________________
Jack-Devel mailing list
Jack-Devel@...
http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org

[Stéphane Letz]

Le 30 oct. 2009 à 11:43, Jörn Nettingsmeier a écrit :

> Stéphane Letz wrote:
>>
>> Le 29 oct. 2009 à 01:11, torbenh a écrit :
>>
>>> this kind of stuff might still happen on osx, because it doesnt have
>>> ppoll, and maybe i need to round up in the wait function.
>>>
>>> anyways, i report slave netxruns with jack_error now. so they should
>>> show up.
>>>
>>>
>>
>>
>> Ok, works OK on 2 Linux machines with default parameter for both  
>> sides
>> of the connections (like "jack_netsbource -h IP and jackd -R -d
>> netone"), but :
>>
>> - I see *3* audio ports on both sides ?? why is that? would be 2 a
>> better default value?
>
> ? 2 *is* the default value.
>
> what you are seeing is two audio ports and one default midi port.

Obviously.. OK

>
>> - changing the port number to 4 or 8 does not works correctly (like
>> "jack_netsbource -h IP -P4 -C4" and  "jackd -R -d netone -i 4 -
>> o4"). I
>> get lot of netxruns on both sides and sound is crappy. (same when
>> changing the latency parameter)
>
> then something is amiss with your setup... are both your endpoints
> low-latency systems? of course in the presence of xruns, the sound is
> "crappy" - that's hardly surprising, is it? thus, no points for
> constructive criticism.
>
> the thing is, you shouldn't be getting xruns, especially not at those
> ridiculously low channel counts.

Tried again on the same setup, worked *once* with 4 channels, then  
again xruns failures with 4 channels and again with default values. I  
tried to start "jack_netsource" first then  "-d netone" or the  
contrary. Is it supposed to make any difference?

I don't see any more meaningful messages when activating "verbose" mode.

Any way to add more helpful messages in the code at least in debug  
mode to understand what happens?

>
> i've been running a 64ch session at 64 frames latency for hours
> yesterday, and it xrunned only when a) there were significant network
> traffic peaks unrelated to netjack, b) one of the endpoints would  
> switch
> cpu frequency, or c) the master would have a hardware xrun.
> it's a best effort thing, no guarantees. but under controlled
> conditions, there were no xruns.

I don't have RT configurated kernel here, but I'm running with a quite  
large (= 1024  frames) buffer size.

>
>> (BTW: why having different naming for *same* parameters on both  
>> side ??
>> (that is  -P -C for in/out channels on master side and -i -o on  
>> slave).
>> It would help very much to have same naming as much as possible...)
>
> yeah, that is something i've been nagging torben about, too. patch is
> underway. but that's cosmetics imho.
>
> let's have netone in the main repo asap, and then we can work on the
> details. my approach would be to eliminate slave configuration
> completely and have jack_netsource dictate all parameters to the  
> slave.
>
> iiuc, the current situation arose because torben has been  
> experimenting
> with runtime reconfiguration of some parameters, and this means the
> slave needs some fixed settings of its own... the final configuration
> approach will evolve over time.
>

Stephane
_______________________________________________
Jack-Devel mailing list
Jack-Devel@...
http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org

[Jörn Nettingsmeier-5]

Stéphane Letz wrote:

>>> - changing the port number to 4 or 8 does not works correctly
>>> (like "jack_netsbource -h IP -P4 -C4" and  "jackd -R -d netone -i
>>> 4 -o4"). I get lot of netxruns on both sides and sound is crappy.
>>> (same when changing the latency parameter)
>>
>> then something is amiss with your setup... are both your endpoints
>> low-latency systems? of course in the presence of xruns, the sound
>> is "crappy" - that's hardly surprising, is it? thus, no points for
>> constructive criticism.
>>
>> the thing is, you shouldn't be getting xruns, especially not at
>> those ridiculously low channel counts.
>
> Tried again on the same setup, worked *once* with 4 channels, then
> again xruns failures with 4 channels and again with default values. I
> tried to start "jack_netsource" first then  "-d netone" or the
> contrary. Is it supposed to make any difference?

don't know. i start "jackd -R -d netone" first, always, but i don't know
whether that's mandatory.

> I don't see any more meaningful messages when activating "verbose"
> mode.
>
> Any way to add more helpful messages in the code at least in debug
> mode to understand what happens?

well, what messages do you get on each side?

what kind of network are you using? can you run gnome-system-monitor or
something similar that will show a history of cpu and network usage
around those xruns?

> I don't have RT configurated kernel here, but I'm running with a
> quite large (= 1024  frames) buffer size.

i must confess i've never tested that, really. 2 days ago, i did try
netone between one rt and one non-rt system with 64 frames/period, and i
had constant xruns. increasing the latency to -l 15 worked. i didn't try
to increase the frame size.



--
Jörn Nettingsmeier

Meister für Veranstaltungstechnik

Audio and event engineer
Ambisonic surround recordings

http://stackingdwarves.net
[phone]

_______________________________________________
Jack-Devel mailing list
Jack-Devel@...
http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org

[Stéphane Letz]

Le 30 oct. 2009 à 17:09, Jörn Nettingsmeier a écrit :

> Stéphane Letz wrote:
>>>> - changing the port number to 4 or 8 does not works correctly
>>>> (like "jack_netsbource -h IP -P4 -C4" and  "jackd -R -d netone -i
>>>> 4 -o4"). I get lot of netxruns on both sides and sound is crappy.
>>>> (same when changing the latency parameter)
>>>
>>> then something is amiss with your setup... are both your endpoints
>>> low-latency systems? of course in the presence of xruns, the sound
>>> is "crappy" - that's hardly surprising, is it? thus, no points for
>>> constructive criticism.
>>>
>>> the thing is, you shouldn't be getting xruns, especially not at
>>> those ridiculously low channel counts.
>>
>> Tried again on the same setup, worked *once* with 4 channels, then
>> again xruns failures with 4 channels and again with default values. I
>> tried to start "jack_netsource" first then  "-d netone" or the
>> contrary. Is it supposed to make any difference?
>
> don't know. i start "jackd -R -d netone" first, always, but i don't  
> know
> whether that's mandatory.
>
>> I don't see any more meaningful messages when activating "verbose"
>> mode.
>>
>> Any way to add more helpful messages in the code at least in debug
>> mode to understand what happens?
>
> well, what messages do you get on each side?

Master:


onnected :-)
at frame 000375 -> total netxruns 1  (0%) queue time= 5324
at frame 000750 -> total netxruns 338  (45%) queue time= 7
at frame 001126 -> total netxruns 676  (60%) queue time= 3
at frame 001501 -> total netxruns 1018  (67%) queue time= 2666
at frame 001877 -> total netxruns 1359  (72%) queue time= 3
at frame 002252 -> total netxruns 1700  (75%) queue time= 8
at frame 002628 -> total netxruns 2043  (77%) queue time= 2
at frame 003004 -> total netxruns 2384  (79%) queue time= 3
at frame 003379 -> total netxruns 2724  (80%) queue time= 2
at frame 003755 -> total netxruns 3066  (81%) queue time= 2
at frame 004131 -> total netxruns 3412  (82%) queue time= 5325
at frame 004507 -> total netxruns 3751  (83%) queue time= 2660
at frame 004882 -> total netxruns 4089  (83%) queue time= 3
at frame 005258 -> total netxruns 4432  (84%) queue time= 3
at frame 005634 -> total netxruns 4776  (84%) queue time= 2657
at frame 006009 -> total netxruns 5119  (85%) queue time= 2660
at frame 006385 -> total netxruns 5457  (85%) queue time= 2
at frame 006760 -> total netxruns 5801  (85%) queue time= 10658
at frame 007136 -> total netxruns 6064  (84%) queue time= 10659
at frame 007511 -> total netxruns 6338  (84%) queue time= 10669


Slave:

netxun... 5886
netxun... 5892


>
> what kind of network are you using? can you run gnome-system-monitor  
> or
> something similar that will show a history of cpu and network usage
> around those xruns?

100 mb network shared with other usages.

(again... works quite well for netjack2 so I don't think the network  
is in cause)

>
>> I don't have RT configurated kernel here, but I'm running with a
>> quite large (= 1024  frames) buffer size.
>
> i must confess i've never tested that, really. 2 days ago, i did try
> netone between one rt and one non-rt system with 64 frames/period,  
> and i
> had constant xruns. increasing the latency to -l 15 worked. i didn't  
> try
> to increase the frame size.
>

Well trying 256 was working much better at some point, then 512  
failing, then working once...

Seems like an init problem or something, since it works from time to  
time.

Again I think some more monitoring code in netjack1 would help better  
understand what happens...

Stéphane

_______________________________________________
Jack-Devel mailing list
Jack-Devel@...
http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org

[Jörn Nettingsmeier-5]

Stéphane Letz wrote:

>
> Le 30 oct. 2009 à 17:09, Jörn Nettingsmeier a écrit :
>
>> Stéphane Letz wrote:
>>>>> - changing the port number to 4 or 8 does not works correctly
>>>>> (like "jack_netsbource -h IP -P4 -C4" and  "jackd -R -d netone -i
>>>>> 4 -o4"). I get lot of netxruns on both sides and sound is crappy.
>>>>> (same when changing the latency parameter)
>>>>
>>>> then something is amiss with your setup... are both your endpoints
>>>> low-latency systems? of course in the presence of xruns, the sound
>>>> is "crappy" - that's hardly surprising, is it? thus, no points for
>>>> constructive criticism.
>>>>
>>>> the thing is, you shouldn't be getting xruns, especially not at
>>>> those ridiculously low channel counts.
>>>
>>> Tried again on the same setup, worked *once* with 4 channels, then
>>> again xruns failures with 4 channels and again with default values. I
>>> tried to start "jack_netsource" first then  "-d netone" or the
>>> contrary. Is it supposed to make any difference?
>>
>> don't know. i start "jackd -R -d netone" first, always, but i don't know
>> whether that's mandatory.
>>
>>> I don't see any more meaningful messages when activating "verbose"
>>> mode.
>>>
>>> Any way to add more helpful messages in the code at least in debug
>>> mode to understand what happens?
>>
>> well, what messages do you get on each side?
>
>
> Master:
>
>
> onnected :-)
> at frame 000375 -> total netxruns 1  (0%) queue time= 5324
> at frame 000750 -> total netxruns 338  (45%) queue time= 7
> at frame 001126 -> total netxruns 676  (60%) queue time= 3
> at frame 001501 -> total netxruns 1018  (67%) queue time= 2666
> at frame 001877 -> total netxruns 1359  (72%) queue time= 3
> at frame 002252 -> total netxruns 1700  (75%) queue time= 8
> at frame 002628 -> total netxruns 2043  (77%) queue time= 2
> at frame 003004 -> total netxruns 2384  (79%) queue time= 3
> at frame 003379 -> total netxruns 2724  (80%) queue time= 2
> at frame 003755 -> total netxruns 3066  (81%) queue time= 2
> at frame 004131 -> total netxruns 3412  (82%) queue time= 5325
> at frame 004507 -> total netxruns 3751  (83%) queue time= 2660
> at frame 004882 -> total netxruns 4089  (83%) queue time= 3
> at frame 005258 -> total netxruns 4432  (84%) queue time= 3
> at frame 005634 -> total netxruns 4776  (84%) queue time= 2657
> at frame 006009 -> total netxruns 5119  (85%) queue time= 2660
> at frame 006385 -> total netxruns 5457  (85%) queue time= 2
> at frame 006760 -> total netxruns 5801  (85%) queue time= 10658
> at frame 007136 -> total netxruns 6064  (84%) queue time= 10659
> at frame 007511 -> total netxruns 6338  (84%) queue time= 10669
>
>
> Slave:
>
> netxun... 5886
> netxun... 5892

those counts are surreal. i mean yeah, if it was an xrun per second, ok.
but this looks more like somebody is wiggling a broken cat5 cable somewhere.


hmm. can you try this:

on the slave:

jackd -R -d netone &


on the master

your usual jackd setup (can you post the commandline?)

jack_netsource -C4 -P4 -i1 -o1 -l1 -h myslave.grame.fr


i'd also be interested in

tracroute master-slave
ping roundtrip time
flood ping using -c 1400
network load

are you using torben's latest tree?

or maybe something is eating fragmented packets?








--
Jörn Nettingsmeier

Meister für Veranstaltungstechnik

Audio and event engineer
Ambisonic surround recordings

http://stackingdwarves.net
[phone]

_______________________________________________
Jack-Devel mailing list
Jack-Devel@...
http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org

[torbenh]

On Fri, Oct 30, 2009 at 05:29:13PM +0100, Stéphane Letz wrote:

>
> Le 30 oct. 2009 à 17:09, Jörn Nettingsmeier a écrit :
>
> >Stéphane Letz wrote:
> >>>>- changing the port number to 4 or 8 does not works correctly
> >>>>(like "jack_netsbource -h IP -P4 -C4" and  "jackd -R -d netone -i
> >>>>4 -o4"). I get lot of netxruns on both sides and sound is crappy.
> >>>>(same when changing the latency parameter)
> >>>
> >>>then something is amiss with your setup... are both your endpoints
> >>>low-latency systems? of course in the presence of xruns, the sound
> >>>is "crappy" - that's hardly surprising, is it? thus, no points for
> >>>constructive criticism.
> >>>
> >>>the thing is, you shouldn't be getting xruns, especially not at
> >>>those ridiculously low channel counts.
> >>
> >>Tried again on the same setup, worked *once* with 4 channels, then
> >>again xruns failures with 4 channels and again with default values. I
> >>tried to start "jack_netsource" first then  "-d netone" or the
> >>contrary. Is it supposed to make any difference?

no... should work fine.

> >
> >don't know. i start "jackd -R -d netone" first, always, but i
> >don't know
> >whether that's mandatory.
> >
> >>I don't see any more meaningful messages when activating "verbose"
> >>mode.
> >>
> >>Any way to add more helpful messages in the code at least in debug
> >>mode to understand what happens?

waht kind of messages do you want ?
shall i print out each received packet ?

> >
> >well, what messages do you get on each side?
>

try running it with -l1

>
> Master:
>
>
> onnected :-)
> at frame 000375 -> total netxruns 1  (0%) queue time= 5324
> at frame 000750 -> total netxruns 338  (45%) queue time= 7
> at frame 001126 -> total netxruns 676  (60%) queue time= 3
> at frame 001501 -> total netxruns 1018  (67%) queue time= 2666
> at frame 001877 -> total netxruns 1359  (72%) queue time= 3
> at frame 002252 -> total netxruns 1700  (75%) queue time= 8
> at frame 002628 -> total netxruns 2043  (77%) queue time= 2
> at frame 003004 -> total netxruns 2384  (79%) queue time= 3
> at frame 003379 -> total netxruns 2724  (80%) queue time= 2
> at frame 003755 -> total netxruns 3066  (81%) queue time= 2
> at frame 004131 -> total netxruns 3412  (82%) queue time= 5325
> at frame 004507 -> total netxruns 3751  (83%) queue time= 2660
> at frame 004882 -> total netxruns 4089  (83%) queue time= 3
> at frame 005258 -> total netxruns 4432  (84%) queue time= 3
> at frame 005634 -> total netxruns 4776  (84%) queue time= 2657
> at frame 006009 -> total netxruns 5119  (85%) queue time= 2660
> at frame 006385 -> total netxruns 5457  (85%) queue time= 2
> at frame 006760 -> total netxruns 5801  (85%) queue time= 10658
> at frame 007136 -> total netxruns 6064  (84%) queue time= 10659
> at frame 007511 -> total netxruns 6338  (84%) queue time= 10669
>
>
> Slave:
>
> netxun... 5886
> netxun... 5892
>
>
> >
> >what kind of network are you using? can you run
> >gnome-system-monitor or
> >something similar that will show a history of cpu and network usage
> >around those xruns?
>
> 100 mb network shared with other usages.
>
> (again... works quite well for netjack2 so I don't think the network
> is in cause)
>
>
> >
> >>I don't have RT configurated kernel here, but I'm running with a
> >>quite large (= 1024  frames) buffer size.
> >
> >i must confess i've never tested that, really. 2 days ago, i did try
> >netone between one rt and one non-rt system with 64 frames/period,
> >and i
> >had constant xruns. increasing the latency to -l 15 worked. i
> >didn't try
> >to increase the frame size.
> >
>
> Well trying 256 was working much better at some point, then 512
> failing, then working once...
>
> Seems like an init problem or something, since it works from time to
> time.
>
> Again I think some more monitoring code in netjack1 would help
> better understand what happens...
>
> Stéphane

--
torben Hohn
http://galan.sourceforge.net -- The graphical Audio language
_______________________________________________
Jack-Devel mailing list
Jack-Devel@...
http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org