[Ninux-Wireless] tinc CPU optimization
Gioacchino Mazzurco
gmazzurco89 a gmail.com
Dom 3 Lug 2011 14:09:07 CEST
altri test fissando la quantita'
[ 4] 0.0-62.2 sec 2.00 MBytes 270 Kbits/sec
[ 4] 0.0-55.3 sec 2.00 MBytes 304 Kbits/sec
[ 4] 0.0-64.2 sec 2.00 MBytes 261 Kbits/sec
[ 4] 0.0-58.8 sec 2.00 MBytes 285 Kbits/sec
[ 4] 0.0-99.6 sec 2.00 MBytes 169 Kbits/sec
[ 4] 0.0-96.4 sec 2.00 MBytes 174 Kbits/sec
[ 4] 0.0-89.8 sec 2.00 MBytes 187 Kbits/sec
[ 4] 0.0-66.4 sec 2.00 MBytes 253 Kbits/sec
[ 4] 0.0-99.9 sec 2.00 MBytes 161 Kbits/sec
[ 4] 0.0-88.1 sec 2.00 MBytes 190 Kbits/sec
Il 03 luglio 2011 13:57, Gioacchino Mazzurco <gmazzurco89 a gmail.com> ha scritto:
> senza tinc la configurazione rimane uguale ma il traffico al posto di
> passare dal tunnel via internet passa solo attraverso i link wireless
>
> Il 03 luglio 2011 13:51, Antonio Quartulli <ordex a autistici.org> ha scritto:
>> On dom, lug 03, 2011 at 01:48:37 +0200, Gioacchino Mazzurco wrote:
>>> il test e' sempre PC( iperf -c ) <-- cavo lan --> Piconstation (
>>> btman-adv + tinc )<-- tinc ---> PC( batman-adv + tinc + iperf -s)
>>
>> anche senza TINC la configurazione rimane uguale? scusa ma non ho capito
>> questo daalle mail precedenti
>>
>>>
>>> >usa un vincolo temporale o quantitativo, sti valori sono troppo
>>> >deviati..
>>>
>>> quei test non sono fatti in parallelo sono fatti in modo sequenziale
>>> quindi volta per volta c'e' ne e' attivo solo uno
>>>
>>> Il 03 luglio 2011 13:40, Darkman <darkman a darkman.it> ha scritto:
>>> > Magari se scegliessi un test "unico" sarebbe anche meglio,
>>> > usa un vincolo temporale o quantitativo, sti valori sono troppo
>>> > deviati..
>>> > Se non mi dicessi della CPU a palla, guardando sta roba ti direi che è
>>> > congestione..
>>> >
>>> > Il giorno 03 luglio 2011 13:31, Gioacchino Mazzurco <gmazzurco89 a gmail.com>
>>> > ha scritto:
>>> >>
>>> >> altra serie di test
>>> >>
>>> >> [ 4] 0.0-18.8 sec 384 KBytes 167 Kbits/sec
>>> >> [ 4] 0.0-17.5 sec 384 KBytes 180 Kbits/sec
>>> >> [ 4] 0.0-20.0 sec 384 KBytes 157 Kbits/sec
>>> >> [ 4] 0.0-21.1 sec 384 KBytes 149 Kbits/sec
>>> >> [ 4] 0.0-23.5 sec 512 KBytes 178 Kbits/sec
>>> >> [ 4] 0.0-32.3 sec 384 KBytes 97.3 Kbits/sec
>>> >> [ 4] 0.0-20.8 sec 384 KBytes 151 Kbits/sec
>>> >> [ 4] 0.0-27.7 sec 256 KBytes 75.8 Kbits/sec
>>> >> [ 4] 0.0-21.8 sec 256 KBytes 96.3 Kbits/sec
>>> >> [ 4] 0.0-14.3 sec 512 KBytes 294 Kbits/sec
>>> >> [ 4] 0.0-14.0 sec 512 KBytes 299 Kbits/sec
>>> >> [ 4] 0.0-37.6 sec 512 KBytes 112 Kbits/sec
>>> >> [ 4] 0.0-18.7 sec 512 KBytes 224 Kbits/sec
>>> >> [ 4] 0.0-21.3 sec 384 KBytes 148 Kbits/sec
>>> >> [ 4] 0.0-17.9 sec 640 KBytes 293 Kbits/sec
>>> >> [ 4] 0.0-24.8 sec 512 KBytes 169 Kbits/sec
>>> >> [ 4] 0.0-23.5 sec 512 KBytes 178 Kbits/sec
>>> >> [ 4] 0.0-16.4 sec 384 KBytes 192 Kbits/sec
>>> >> [ 4] 0.0-21.4 sec 384 KBytes 147 Kbits/sec
>>> >>
>>> >> ho spento dnsmasq che non serviva a niente e andiamo di poco ma meglio
>>> >>
>>> >> Il 03 luglio 2011 13:16, Darkman <darkman a darkman.it> ha scritto:
>>> >> > Il sintomo è abbastanza chiaro, ma dubito sia colpa della CPU o meglio,
>>> >> > secondo me qualcosa
>>> >> > è stata scritta male, 100Kbps sono davvero ridicoli. A maggior ragione
>>> >> > quando ste cpu hanno anche qualche set dedicato
>>> >> > alla crittografia simmetrica...
>>> >> >
>>> >> > Il giorno 03 luglio 2011 13:04, Gioacchino Mazzurco
>>> >> > <gmazzurco89 a gmail.com>
>>> >> > ha scritto:
>>> >> >>
>>> >> >> ma il problema sembra proprio l'eccessivo utilizzo di cpu per la vpn
>>> >> >> perche' stando in ssh sulla picostation mentre c'e' traffico che passa
>>> >> >> sulla vpn diventa completamente unresponsive non sente nemmeno ctrl+c
>>> >> >> sulla shell... quando il traffico finisce mi esegue tutto quello che
>>> >> >> gli avevo mandato nel fratempo
>>> >> >>
>>> >> >> Il 03 luglio 2011 13:01, Gioacchino Mazzurco <gmazzurco89 a gmail.com> ha
>>> >> >> scritto:
>>> >> >> >>Hai la possibilità di usare una CPU + potente (tincare dal PC)?
>>> >> >> >
>>> >> >> > dovrei installarmi anche batman-adv sul pc...
>>> >> >> >
>>> >> >> > Il 03 luglio 2011 12:58, Darkman <darkman a darkman.it> ha scritto:
>>> >> >> >> E' chiaro che non può essere il tuo upstream,
>>> >> >> >> ma sei certo che il collo di bottiglia non sia nella capacità di sta
>>> >> >> >> rete
>>> >> >> >> mesh tunnellata?
>>> >> >> >> Hai provato a lanciare 2 iperf in parallelo?
>>> >> >> >> Hai la possibilità di usare una CPU + potente (tincare dal PC)?
>>> >> >> >>
>>> >> >> >>
>>> >> >> >> Il giorno 03 luglio 2011 12:34, Gioacchino Mazzurco
>>> >> >> >> <gmazzurco89 a gmail.com>
>>> >> >> >> ha scritto:
>>> >> >> >>>
>>> >> >> >>> la picostation a e la z sono la stessa picostation... dalla
>>> >> >> >>> picostation a posso decidere se accendere tinc e quindi far passare
>>> >> >> >>> traffico mesh su internet oppure se usare solo i link wireless
>>> >> >> >>>
>>> >> >> >>> dal computer pocco decidere sia di usare la picostation come gw sia
>>> >> >> >>> di
>>> >> >> >>> usare il router adsl
>>> >> >> >>>
>>> >> >> >>> le casistiche quindi sono 3
>>> >> >> >>>
>>> >> >> >>> iperf via internet senza tinc >500KB/s
>>> >> >> >>> iperf via mesh senza tinc ~ 20Kb/s
>>> >> >> >>> iperf via mesh tunnellata su internet con tinc ~100Kb/s
>>> >> >> >>>
>>> >> >> >>> Il 03 luglio 2011 12:27, Darkman <darkman a darkman.it> ha scritto:
>>> >> >> >>> > Fammi capire:
>>> >> >> >>> > - tra le tua pico(A) e quella(Z) con l'adsl ci sono diversi nodi
>>> >> >> >>> > e
>>> >> >> >>> > con
>>> >> >> >>> > iperf
>>> >> >> >>> > hai risultati di 20Kbps (A->Z) in L3 puro ? Mentre se usi tinc va
>>> >> >> >>> > a
>>> >> >> >>> > 100Kbps?
>>> >> >> >>> > - chi sono gli end-point tinc?
>>> >> >> >>> >
>>> >> >> >>> >
>>> >> >> >>> >
>>> >> >> >>> >
>>> >> >> >>> >
>>> >> >> >>> > Il giorno 03 luglio 2011 12:12, Gioacchino Mazzurco
>>> >> >> >>> > <gmazzurco89 a gmail.com>
>>> >> >> >>> > ha scritto:
>>> >> >> >>> >>
>>> >> >> >>> >> senza tinc praticamente non c'e' connettivita' ( a volte va ma
>>> >> >> >>> >> roba
>>> >> >> >>> >> tipo 20k perche' sono un sacco di op alcuni dei quali fanno
>>> >> >> >>> >> schifo...)
>>> >> >> >>> >>
>>> >> >> >>> >> se invece faccio iperf passando per internet senza tinc ottengo
>>> >> >> >>> >> risultati sempre sopra i 500KB/s
>>> >> >> >>> >>
>>> >> >> >>> >> Il 03 luglio 2011 12:01, Darkman <darkman a darkman.it> ha
>>> >> >> >>> >> scritto:
>>> >> >> >>> >> > Hai gia controllato i valori tra le 2 pico con e senza tinc?
>>> >> >> >>> >> >
>>> >> >> >>> >> >
>>> >> >> >>> >> > Il giorno 03 luglio 2011 11:45, Gioacchino Mazzurco
>>> >> >> >>> >> > <gmazzurco89 a gmail.com>
>>> >> >> >>> >> > ha scritto:
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> iperf -c su computer che usa una picostation come gateway ->
>>> >> >> >>> >> >> Picostation con tinc <- adsl 8 megabit -> iperf --server su
>>> >> >> >>> >> >> eigenlab.org
>>> >> >> >>> >> >>
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> Il 03 luglio 2011 11:33, Darkman <darkman a darkman.it> ha
>>> >> >> >>> >> >> scritto:
>>> >> >> >>> >> >> > 100kbps mi pare davvero troppo poco anche per quelle
>>> >> >> >>> >> >> > cessonanocpu.
>>> >> >> >>> >> >> > Come li hai ottenuti sti valori?
>>> >> >> >>> >> >> >
>>> >> >> >>> >> >> > Il giorno 03 luglio 2011 11:10, Gioacchino Mazzurco
>>> >> >> >>> >> >> > <gmazzurco89 a gmail.com>
>>> >> >> >>> >> >> > ha scritto:
>>> >> >> >>> >> >> >>
>>> >> >> >>> >> >> >> Ciao a tutti!
>>> >> >> >>> >> >> >>
>>> >> >> >>> >> >> >> Facendo dei test mi sono accorto che le vpn con tinc
>>> >> >> >>> >> >> >> installato
>>> >> >> >>> >> >> >> sui
>>> >> >> >>> >> >> >> nodi ci vanno max a 100k anche se la banda dell'adsl e'
>>> >> >> >>> >> >> >> molta
>>> >> >> >>> >> >> >> di
>>> >> >> >>> >> >> >> piu'... ho cominciato a cercare ed ho letto che la causa
>>> >> >> >>> >> >> >> e'
>>> >> >> >>> >> >> >> probabilmente la CPU che non ce la fa a fare encryption
>>> >> >> >>> >> >> >> decryption
>>> >> >> >>> >> >> >> piu' velocemente di cosi'
>>> >> >> >>> >> >> >>
>>> >> >> >>> >> >> >> leggendo il man di tinc ho trovato questo
>>> >> >> >>> >> >> >>
>>> >> >> >>> >> >> >> Cipher = cipher (blowfish)
>>> >> >> >>> >> >> >> The symmetric cipher algorithm used to encrypt
>>> >> >> >>> >> >> >> UDP
>>> >> >> >>> >> >> >> packets. Any cipher supported by OpenSSL is recognised.
>>> >> >> >>> >> >> >> Furâ€
>>> >> >> >>> >> >> >> thermore, specifying "none" will turn off
>>> >> >> >>> >> >> >> packet
>>> >> >> >>> >> >> >> encryption. It is best to use only those ciphers which
>>> >> >> >>> >> >> >> support
>>> >> >> >>> >> >> >> CBC mode.
>>> >> >> >>> >> >> >>
>>> >> >> >>> >> >> >> mettendo none dovrebbe essere disabilitata l' encryption e
>>> >> >> >>> >> >> >> quindi
>>> >> >> >>> >> >> >> avere piu' banda, il meccanismo degli host con il file con
>>> >> >> >>> >> >> >> la
>>> >> >> >>> >> >> >> chiave
>>> >> >> >>> >> >> >> pubblica continua a funzionare disabilitando la cifratura,
>>> >> >> >>> >> >> >> e
>>> >> >> >>> >> >> >> soprattutto bastera' aggiungere quell'opzione li oppure
>>> >> >> >>> >> >> >> bisogna
>>> >> >> >>> >> >> >> cambiare altre conf?
>>> >> >> >>> >> >> >>
>>> >> >> >>> >> >> >> _______________________________________________
>>> >> >> >>> >> >> >> Wireless mailing list
>>> >> >> >>> >> >> >> Wireless a ml.ninux.org
>>> >> >> >>> >> >> >> http://ml.ninux.org/mailman/listinfo/wireless
>>> >> >> >>> >> >> >>
>>> >> >> >>> >> >> >
>>> >> >> >>> >> >> >
>>> >> >> >>> >> >> > _______________________________________________
>>> >> >> >>> >> >> > Wireless mailing list
>>> >> >> >>> >> >> > Wireless a ml.ninux.org
>>> >> >> >>> >> >> > http://ml.ninux.org/mailman/listinfo/wireless
>>> >> >> >>> >> >> >
>>> >> >> >>> >> >> >
>>> >> >> >>> >> >> _______________________________________________
>>> >> >> >>> >> >> Wireless mailing list
>>> >> >> >>> >> >> Wireless a ml.ninux.org
>>> >> >> >>> >> >> http://ml.ninux.org/mailman/listinfo/wireless
>>> >> >> >>> >> >>
>>> >> >> >>> >> >
>>> >> >> >>> >> >
>>> >> >> >>> >> > _______________________________________________
>>> >> >> >>> >> > Wireless mailing list
>>> >> >> >>> >> > Wireless a ml.ninux.org
>>> >> >> >>> >> > http://ml.ninux.org/mailman/listinfo/wireless
>>> >> >> >>> >> >
>>> >> >> >>> >> >
>>> >> >> >>> >> _______________________________________________
>>> >> >> >>> >> Wireless mailing list
>>> >> >> >>> >> Wireless a ml.ninux.org
>>> >> >> >>> >> http://ml.ninux.org/mailman/listinfo/wireless
>>> >> >> >>> >>
>>> >> >> >>> >
>>> >> >> >>> >
>>> >> >> >>> > _______________________________________________
>>> >> >> >>> > Wireless mailing list
>>> >> >> >>> > Wireless a ml.ninux.org
>>> >> >> >>> > http://ml.ninux.org/mailman/listinfo/wireless
>>> >> >> >>> >
>>> >> >> >>> >
>>> >> >> >>> _______________________________________________
>>> >> >> >>> Wireless mailing list
>>> >> >> >>> Wireless a ml.ninux.org
>>> >> >> >>> http://ml.ninux.org/mailman/listinfo/wireless
>>> >> >> >>>
>>> >> >> >>
>>> >> >> >>
>>> >> >> >> _______________________________________________
>>> >> >> >> Wireless mailing list
>>> >> >> >> Wireless a ml.ninux.org
>>> >> >> >> http://ml.ninux.org/mailman/listinfo/wireless
>>> >> >> >>
>>> >> >> >>
>>> >> >> >
>>> >> >> _______________________________________________
>>> >> >> Wireless mailing list
>>> >> >> Wireless a ml.ninux.org
>>> >> >> http://ml.ninux.org/mailman/listinfo/wireless
>>> >> >>
>>> >> >
>>> >> >
>>> >> > _______________________________________________
>>> >> > Wireless mailing list
>>> >> > Wireless a ml.ninux.org
>>> >> > http://ml.ninux.org/mailman/listinfo/wireless
>>> >> >
>>> >> >
>>> >> _______________________________________________
>>> >> Wireless mailing list
>>> >> Wireless a ml.ninux.org
>>> >> http://ml.ninux.org/mailman/listinfo/wireless
>>> >>
>>> >
>>> >
>>> > _______________________________________________
>>> > Wireless mailing list
>>> > Wireless a ml.ninux.org
>>> > http://ml.ninux.org/mailman/listinfo/wireless
>>> >
>>> >
>>> _______________________________________________
>>> Wireless mailing list
>>> Wireless a ml.ninux.org
>>> http://ml.ninux.org/mailman/listinfo/wireless
>>
>> --
>> Antonio Quartulli
>>
>> ..each of us alone is worth nothing..
>> Ernesto "Che" Guevara
>> _______________________________________________
>> Wireless mailing list
>> Wireless a ml.ninux.org
>> http://ml.ninux.org/mailman/listinfo/wireless
>>
>
Maggiori informazioni sulla lista
Wireless