[Cerowrt-devel] fq_codel through Tor

[Cerowrt-devel] fq_codel through Tor

[Dave Taht]

[-- Attachment #1: Type: text/plain, Size: 296 bytes --]

https://srv1.openwireless.org/pipermail/tech/2012-December/000332.html

I haven't the foggiest idea what this traffic would look like. Is it even
possible to induce bufferbloat through tor?

-- 
Dave Täht

Fixing bufferbloat with cerowrt:
http://www.teklibre.com/cerowrt/subscribe.html

[-- Attachment #2: Type: text/html, Size: 518 bytes --]

Re: [Cerowrt-devel] fq_codel through Tor

[David Lang]

[-- Attachment #1: Type: TEXT/Plain, Size: 614 bytes --]

On Sat, 19 Jan 2013, Dave Taht wrote:

> https://srv1.openwireless.org/pipermail/tech/2012-December/000332.html
>
> I haven't the foggiest idea what this traffic would look like. Is it even
> possible to induce bufferbloat through tor?

with your packets bouncing through so many systems there are sure a lot of 
buffers in play. Especially given that Tor is not trying to find the most 
efficent route to send the packets, I would not be surprised to see occasional 
bottlenecks.

On the other hand, I don't know what you can do other than the basic queue 
management stuff you are already working on.

David Lang

[-- Attachment #2: Type: TEXT/PLAIN, Size: 164 bytes --]

_______________________________________________
Cerowrt-devel mailing list
Cerowrt-devel@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/cerowrt-devel

Re: [Cerowrt-devel] fq_codel through Tor

[Maciej Soltysiak]

[-- Attachment #1: Type: text/plain, Size: 1313 bytes --]

Funny you should ask. Being inspired by Tor's Jacob Applebaum's keynote at
#29C3 ( http://isoc-ny.org/p2/4650 ) I started a tor node. Without
throttling the effect on my box was similar to bittorrent : instantly
dozens of connections consuming in total 4-5 MB/s inwards and outwards.
Observed in iptraf. ssh felt a bit laggy.

I think much depends on your exit a policy. If you allow all no port
restrictions (default) you might be serving a lot, perhaps even bit torrent;

I saw a headline somewhere about ways to circumvent tor policy to run
torrents.

SO unbloated devices may be keen on unbloating to still live with being
generous to tor which is very important for the project as the main issue
with it is it's slowness.

Maciej
On 19 Jan 2013 09:57, "Dave Taht" <dave.taht@gmail.com> wrote:

> https://srv1.openwireless.org/pipermail/tech/2012-December/000332.html
>
> I haven't the foggiest idea what this traffic would look like. Is it even
> possible to induce bufferbloat through tor?
>
> --
> Dave Täht
>
> Fixing bufferbloat with cerowrt:
> http://www.teklibre.com/cerowrt/subscribe.html
> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>
>

[-- Attachment #2: Type: text/html, Size: 2037 bytes --]

Re: [Cerowrt-devel] fq_codel through Tor

[dpreed]

[-- Attachment #1: Type: text/plain, Size: 2753 bytes --]


Tor should be following the same rules as routers - buffer minimally, signal congestion quickly (by packet drop, ECN, etc. on an end-to-end basis).  I bet it does *not* do the latter at its layer, and I bet the underlying (non Tor) layer does not.
 
Remember - using fq_codel in a "home router" does not fix the real problem in the DOCSIS deployment, nor does it fix the real problem in the LTE deployment.  By fixing the "outgoing rate" less than the "service rate", you just never use (hopefully) the buffers in your cable modem uplink, which are not shared with other users.
 
But Tor is a system of "routers" (onion-y ones), and its own "software" needs to be fixed.
 
Is anyone actually fixing the Tor router layer?
 
That's not sufficient, because the layer 2 below the IP layer *under* Tor will still be bad.
 
But it may not be worth fixing one layer without fixing the other.
 
Tor's buffering should be studied.
 
-----Original Message-----
From: "Maciej Soltysiak" <maciej@soltysiak.com>
Sent: Saturday, January 19, 2013 5:03am
To: "Dave Taht" <dave.taht@gmail.com>
Cc: ju@klipix.org, cerowrt-devel@lists.bufferbloat.net
Subject: Re: [Cerowrt-devel] fq_codel through Tor



Funny you should ask. Being inspired by Tor's Jacob Applebaum's keynote at #29C3 ( [http://isoc-ny.org/p2/4650] http://isoc-ny.org/p2/4650 ) I started a tor node. Without throttling the effect on my box was similar to bittorrent : instantly dozens of connections consuming in total 4-5 MB/s inwards and outwards. Observed in iptraf. ssh felt a bit laggy.
I think much depends on your exit a policy. If you allow all no port restrictions (default) you might be serving a lot, perhaps even bit torrent;
I saw a headline somewhere about ways to circumvent tor policy to run torrents.
SO unbloated devices may be keen on unbloating to still live with being generous to tor which is very important for the project as the main issue with it is it's slowness.
Maciej
On 19 Jan 2013 09:57, "Dave Taht" <[mailto:dave.taht@gmail.com] dave.taht@gmail.com> wrote:
[https://srv1.openwireless.org/pipermail/tech/2012-December/000332.html] https://srv1.openwireless.org/pipermail/tech/2012-December/000332.html

I haven't the foggiest idea what this traffic would look like. Is it even possible to induce bufferbloat through tor?-- 
Dave Täht

Fixing bufferbloat with cerowrt: [http://www.teklibre.com/cerowrt/subscribe.html] http://www.teklibre.com/cerowrt/subscribe.html 
_______________________________________________
 Cerowrt-devel mailing list
[mailto:Cerowrt-devel@lists.bufferbloat.net] Cerowrt-devel@lists.bufferbloat.net
[https://lists.bufferbloat.net/listinfo/cerowrt-devel] https://lists.bufferbloat.net/listinfo/cerowrt-devel


[-- Attachment #2: Type: text/html, Size: 4009 bytes --]

Re: [Cerowrt-devel] fq_codel through Tor

[Dave Taht]

[-- Attachment #1: Type: text/plain, Size: 3818 bytes --]

On Sat, Jan 19, 2013 at 8:39 AM, <dpreed@reed.com> wrote:

> Tor should be following the same rules as routers - buffer minimally,
> signal congestion quickly (by packet drop, ECN, etc. on an end-to-end
> basis).  I bet it does *not* do the latter at its layer, and I bet the
> underlying (non Tor) layer does not.
>
>
>
> Remember - using fq_codel in a "home router" does not fix the real problem
> in the DOCSIS deployment, nor does it fix the real problem in the LTE
> deployment.  By fixing the "outgoing rate" less than the "service rate",
> you just never use (hopefully) the buffers in your cable modem uplink,
> which are not shared with other users.
>
>
>
> But Tor is a system of "routers" (onion-y ones), and its own "software"
> needs to be fixed.
>
>
>
> Is anyone actually fixing the Tor router layer?
>
>
>
> That's not sufficient, because the layer 2 below the IP layer *under* Tor
> will still be bad.
>
>
>
> But it may not be worth fixing one layer without fixing the other.
>
>
>
> Tor's buffering should be studied.
>

Agreed. Tor qualifies as pretty interesting traffic. It would be my hope,
however, that with a rate shaper on and fq_codel on top of that, that you'd
hardly notice it was there, if you were acting as a relay, and if you used
it yourself, except for laggy things like dns.

In looking through their full transparent proxy stuff,

https://trac.torproject.org/projects/tor/wiki/doc/TransparentProxy

It looks like that, and polipo and the tor build I already have in cero is
enough to do some serious fiddling with tor

https://trac.torproject.org/projects/tor/wiki/doc/CentralizedTorServer

Hmm. I wonder what TFO does to tor's notions...

Anyway:

I'm not terribly interested in routing all traffic through it personally,
but I wouldn't mind setting up a redirect for .onion domains,
 and routing .onion through tor to (for example) setup test web servers on
various sides of the onion to observe what happens.

and...

Can something like netperf get run through it?




>
> -----Original Message-----
> From: "Maciej Soltysiak" <maciej@soltysiak.com>
> Sent: Saturday, January 19, 2013 5:03am
> To: "Dave Taht" <dave.taht@gmail.com>
> Cc: ju@klipix.org, cerowrt-devel@lists.bufferbloat.net
> Subject: Re: [Cerowrt-devel] fq_codel through Tor
>
>  Funny you should ask. Being inspired by Tor's Jacob Applebaum's keynote
> at #29C3 ( http://isoc-ny.org/p2/4650 ) I started a tor node. Without
> throttling the effect on my box was similar to bittorrent : instantly
> dozens of connections consuming in total 4-5 MB/s inwards and outwards.
> Observed in iptraf. ssh felt a bit laggy.
>
> I think much depends on your exit a policy. If you allow all no port
> restrictions (default) you might be serving a lot, perhaps even bit torrent;
>
> I saw a headline somewhere about ways to circumvent tor policy to run
> torrents.
>
> SO unbloated devices may be keen on unbloating to still live with being
> generous to tor which is very important for the project as the main issue
> with it is it's slowness.
>
> Maciej
> On 19 Jan 2013 09:57, "Dave Taht" <dave.taht@gmail.com> wrote:
>
>> https://srv1.openwireless.org/pipermail/tech/2012-December/000332.html
>> I haven't the foggiest idea what this traffic would look like. Is it even
>> possible to induce bufferbloat through tor?
>> --
>> Dave Täht
>>
>> Fixing bufferbloat with cerowrt:
>> http://www.teklibre.com/cerowrt/subscribe.html
>> _______________________________________________
>> Cerowrt-devel mailing list
>> Cerowrt-devel@lists.bufferbloat.net
>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>>
>>


-- 
Dave Täht

Fixing bufferbloat with cerowrt:
http://www.teklibre.com/cerowrt/subscribe.html

[-- Attachment #2: Type: text/html, Size: 6431 bytes --]