-----Original Message----- From: tor-relays [mailto:tor-relays-bounces@lists.torproject.org] On Behalf Of Sebastian Niehaus Sent: Tuesday, December 13, 2016 9:07 PM To: tor-relays@lists.torproject.org Subject: Re: [tor-relays] Tor relay from home - end of experiment? Am 13.12.2016 um 20:01 schrieb Rana:

...

Any other advice / ideas welcome.

You have been asked for fingerpring or atlas link several times.

The nicknames of the two relays are ZG0 and GG2, respectively

-----Original Message-----

From: tor-relays [mailto:tor-relays-bounces@lists.torproject.org] On Behalf Of teor

Your relay's observed bandwidth is ZG0: 81.35 KByte/s GG2: 170.79 KByte/s

(hover over the bandwidth heading in atlas for these details), so its consensus weight will be limited to approximately ZG0: 81 GG2: 170 Your bandwidth rate and burst are ZG0: 256 KByte/s and 358.4 KByte/s GG2: 153.6 KByte/s and 179.2 KByte/s

Your relay is limited by: ZG0: its own ability to sustain more than 81.35 KByte/s over a 10 second period GG2: the bandwidth rate of 153.6 KByte/s

In summary: ZG0: the relay itself reports that it is unable to sustain much tor traffic. GG2: it appears that the relay could handle more traffic, if you increased the bandwidth rate and bandwidth burst.

Both ZG0 and GG2 use exactly the same hardware and software (Pi with the same microsd card flashed), with the exception that ZG0's max bandwidth rates are actually configured to be HIGHER than those of GG2. Logical conclusion based on the above measurements: ZG0's poor bandwidth has absolutely nothing to do with the capabilities of ZG0 machine to sustain traffic. This can only have anything to do with its ISP and/or with the way the Authorities "treat" ZG0. Note that on both machines cpu utilization is minuscule and memory utilization is 12%. Why, while GG2's speed as MEASURED by tor and reported in atlas is 153.6 kbytes/sec, its actual bandwidth utilization is about 0.3 kbytes/sec (=0.2% of its capability) is still beyond me, and resolving this by further increasing bandwidth does not seem plausible to me.

I feel like I've given the same advice about ZG0 several times now, so I'm going to leave that with you to resolve however you want.

You and some others have been very helpful in educating me about what is going on, and for that I thank you. However, is described above, I am not anywhere near resolving the issue. The only conclusion is that I simply cannot run relays from home locations, because their use will be negligible, whatever I (reasonably) do. This opinion was further supported by the reported two experiments with two identical Pi-based relay instances running at two different home locations with two different ISPs. Among other things, I was given (I do not remember by whom) the explanation that the bandwidth ratings of my relays suck probably because they are measured from where the Authorities are located, and my connection to THESE locations may not be good. This may well be the correct explanation that reaffirms my conclusion that all I can do with my two relays is ditch them. I want to reiterate my opinion that Tor network is "mistreating" home-based relays without good reason: A. The fact that the Authorities are located in West Europe and North America does not mean that the USERS are there. I would suggest quite the opposite: the users that REALLY need Tor are NOT located in these countries. Bandwidth measurements should be performed, among others, from where the most needful users are located, not only from democratic countries that host DirAuths. While I understand why DirAuths need to be located in safe places, I see no reason for bwauths that measure the bandwidth to only be located there. DirAuths are not disposable, bwauths are. B. There are about 7000 relays total, many of them probably limping just like my 2 relays and not being useful. There are tens of thousands of Pi owners who have their Pis just sitting there and many of them would be happy to run relays if Tor network would let them do so usefully. Not using this huge resource by discriminating against relays that are behind dynamic IPs or because they happen to have a poor connection to, say, Germany, does not make sense to me. Rana

On 14 Dec. 2016, at 19:47, Rana <ranaventures@gmail.com> wrote:

Why, while GG2's speed as MEASURED by tor and reported in atlas is 153.6 kbytes/sec, its actual bandwidth utilization is about 0.3 kbytes/sec (=0.2% of its capability) is still beyond me, and resolving this by further increasing bandwidth does not seem plausible to me.

The observed bandwidth is the *maximum* (highest, largest, most) bandwidth your relay has sustained over the most utilised (biggest, busiest) 10 second period in the last 24? hours. https://en.wikipedia.org/wiki/Sample_maximum_and_minimum The bandwidth utilisation you quote is the *average* (mean, overall, evened out) bandwidth your relay has used over the last 24? hours. https://en.wikipedia.org/wiki/Arithmetic_mean If you don't understand the difference between a maximum and an average, perhaps it would help to read the links above. Again, I'll leave that with you.

On 14 Dec. 2016, at 20:36, Andreas Krey <a.krey@gmx.de> wrote:

...

A. The fact that the Authorities are located in West Europe and North America does not mean that the USERS are there.

That does not matter - they themselves are well connected, and measure bandwidth, not ping times. It might just be that home dsl providers have bad peering, as rumoured for german telekom and some north american providers. Putting bandwidth auths behind some net curtain would optimize the bandwidth measurements for that specific curtain, which would not help people behind other curtains with different holes/peerings. The question is what volume a relay can carry, and not how well it is connected to a particular place in the world.

The bwauth calculations do take latency into account, and they should: if CPU usage or bandwidth are near their limit, the latency through the relay will be high. (Or connections will fail: https://gitweb.torproject.org/torflow.git/tree/NetworkScanners/BwAuthority/d... ) This has the drawback that relays located away from the US/Western Europe get poor scores. There's no way to compensate for this without it being gamed, except for distributing the bandwidth authorities closer to either relays or users. But this has a drawback: the Tor network is fast because the consensus weight centroid is located in an area which has a lot of low-latency connections. If we shift that centroid, then the network will slow down. This is a problem that needs more research :-) T -- Tim Wilson-Brown (teor) teor2345 at gmail dot com PGP C855 6CED 5D90 A0C5 29F6 4D43 450C BA7F 968F 094B ricochet:ekmygaiu4rzgsk6n xmpp: teor at torproject dot org ------------------------------------------------------------------------

On 14 Dec. 2016, at 22:42, Andreas Krey <a.krey@gmx.de> wrote:

On Wed, 14 Dec 2016 21:43:28 +0000, teor wrote: ...

...

The bwauth calculations do take latency into account, and they should: if CPU usage or bandwidth are near their limit, the latency through the relay will be high.

I stand corrected.

I observed my relays (a few years ago) to often run into the bandwidth limit, aka 'flatlining', and this having latency. I then started to set lower advertised bandwidth, and this went away. Problem here is that these are short-term event in relation to the bandwidth probes, so the probing can't really control this.

...

...

This has the drawback that relays located away from the US/Western Europe get poor scores.

What kind of latencies are we talking about here? And how much latency makes up for what bandwidth?

That's a really good question, the following factors interact: * bandwidth, * network peering, * latency. I don't know the answer, but in my experience, an Exit relay located in France picked up bandwidth extremely quickly (peak in 4-5 months), one in Canada was a little slower (6 months and still at 60% or so) and some non-Exit relays in Australia are operating at 10% of capacity. (They don't all have the same bandwidth, so it's not a fair comparison.) I'd have to ping each of the bwauths from each relay to be sure of the latencies, but as an example: To moria1: France: 94 ms Australia: 245 ms To gabelmoobwscan: France: 15 ms Australia: 340 ms Note that it takes several round-trips to set up a TCP connection, and more round-trips to set up a Tor circuit (and acknowledge cells). T -- Tim Wilson-Brown (teor) teor2345 at gmail dot com PGP C855 6CED 5D90 A0C5 29F6 4D43 450C BA7F 968F 094B ricochet:ekmygaiu4rzgsk6n xmpp: teor at torproject dot org ------------------------------------------------------------------------

...

A. The fact that the Authorities are located in West Europe and North America does not mean that the USERS are there.

The question is what volume a relay can carry, and not how well it is connected to a particular place in the world.

I beg to differ. My experiment with two identical Pies in the same country showed that the alleged volume that the relay can carry IS dependent on how well it is connected to the specific DirAuths (which represent "particular places in the world"). The fact is, the two nodes are HW/SW-wise identical and both have much more Internet connection bandwidth than the bandwidth allocated for Tor. They do, however, have different numbers as to how much traffic they can carry; which in view of the above IMHO can be attributed only to the difference in how well their respective IPSs connect with the ISPs in places where DirAuths are located.

...

B. There are about 7000 relays total, many of them probably limping just like my 2 relays and not being useful. There are tens of thousands of Pi owners who have their Pis just sitting there and many >>of them would be happy to run relays if Tor network would let them do so usefully.

I may soon have an opportunity to hook up a pi to a sufficiently large pipe. (My home connection makes such things pointless.)

Your upcoming connection of a Pi to a large pipe is irrelevant to the issue reported by me, since clearly in my case the Pi is not the bottleneck. On the other hand, your parenthesized sentence is very relevant - it seems that you have given up on home based relay, too. I did see a report from someone boasting the large bandwidth via Pi at home - but this seems to be an exception rather than a rule, and he was in Germany, probably at a cozy digital distance from the local DirAuth :)

Since the consensus weight is the low-median of 5 measurements spread around the US and Western Europe, being in Germany only gets you one good measurement: you need 3 good measurements >to get a high consensus weight.

From https://en.wikipedia.org/wiki/Median :

Because of this, the median is of central importance in robust statistics, as it is the most resistant statistic, having a breakdown point of 50%: so long as no more than half the data are contaminated, the median will not give an arbitrarily large or small result.

I know what average, median and peak is, thank you for the lecture in high school math. The difference in the definitions of peak and average cannot possibly account for the latter being 0.2% of the former in GG2, this ratio being pretty stable over the last 100 hours.

Am 14.12.2016 um 11:46 schrieb Rana:

They do, however, have different numbers as to how much traffic they can carry; which in view of the above IMHO can be attributed only to the difference in how well their respective IPSs connect with the ISPs in places where DirAuths are located.

What kind of connection are they attached to? DSL? Fiber? Cable? POTS/Modem? ... ? What kind of hardware is used to build the connection? One of these shiny pretty Netgear R7900? Sebastian

On 14 Dec 2016 14:49, "Rana" <ranaventures@gmail.com> wrote: -----Original Message----- From: tor-relays [mailto:tor-relays-bounces@lists.torproject.org] On Behalf Of Sebastian Niehaus Sent: Wednesday, December 14, 2016 2:43 PM To: tor-relays@lists.torproject.org Subject: Re: [tor-relays] Tor relay from home - end of experiment? Am 14.12.2016 um 11:46 schrieb Rana:

...

They do, however, have different numbers as to how much traffic they can carry; which in view of the above IMHO can be attributed only to the difference in how well their respective IPSs connect with the ISPs in places where DirAuths are located. What kind of connection are they attached to? DSL? Fiber? Cable? POTS/Modem? ... ? DSL, both of them - but different ISPs. ZG0 has up bandwidth of 1.5mbps, GG2 has 3mbps, both very table at these numbers

What kind of hardware is used to build the connection? One of these shiny pretty Netgear R7900? DSL modem boxes supplied by the ISPs. In case of ZG0, an ADB box. Why is this significant?

Some are quite limited in the number of simultaneous connections they can handle for example (see complaints from bittorrent users), but I doubt this is the case here as you don't get many users