Very few devices can do true geo, which isn’t what they’re trying to do… since geo has minimal bearing on network performance. Most existing geo methods don’t get much better than what country you’re in.

A Comcast user in Chicago probably has a better connection to a Comcast user in Seattle than the AT&T user in Chicago.

People, you will never be anonymous on the Internet, so give it up. It’s impossible. Think of a letter. Someone mails it to your house, you get caught. You give them your friend’s address and he hands it to you. Your friend gets caught. How many friends do you think it’ll take before none will let you have letters mailed to their house? I’m not saying that privacy is a bad thing, but it is an impossibility when you are receiving something… unless you’re /dev/null.

It also requires 0 ISP involvement.

Am I an ISP? Yes.

Most if not all of the ISPs aren’t around to bork you in the bum. There are immense financial and technical reasons why P2P is throttled. Some are more aggressive, but something that benefits both sides is a good thing.

Nice work thank you. More speed and more versions about your software is needed to improve speed and remove bugs.

So while your proximity biasing approach might ease the load on ISPs
peering costs, the consensus is that throughput depends on many factors
like the amount of bandwidth your connected peers are willing to
dedicate to you, the congestion level of your peers and so on.

An approach I now chose with the no-seeders-for-seeders problem is based
on the observation that 200 peers in a tracker reply almost always is
enough to get enough leechers when trying to seed. However the client
doesn’t know who is – until it connects. The tracker OTOH does know.

Instead of implementing a filter on the tracker side – that might run
into the same problems as your proposal does – I chose to just provide
the missing information to the clients (in form of a bit field) and let
them chose what to do.

Now for you there’s several ways to move on. You can throw money on the
problem and just buy a cluster huge enough to handle the load – when I
fixed the multithreading problems in the underlying IO-lib, more cores
will also help.

You could also resort to doing what I do with the seeders flag: just
return random peers but pass the information to the client (4bit
“distance” info will increase reply by 100 bytes max.) and let clients
decide whom to connect to. This results in 200 lookups per announce max
and reduces complexity to O(1) again.

You could also chose a more subtle approach by re-rolling the dice in
each sliding window positions: If the first choice’s distance is above a
certain threshold, you try m others and then chose the closest of the
peers. This still gives you O(1*m). Plus you can fine tune the threshold
and amount of re-tries.

For day to day use on our _open_ opentracker installations however this
biasing code feels a little like betraying our peers. Outlandish peers
suddenly will experience degradations in their p2p quality. Applying
some magic filtering code to exclude certain peers from replies feels
like a wrong thing to do on an open tracker. OTOH client authors say
that any pre-filtering on the tracker is void after they connect to the
first peer and receive more peers via PEX and DHT.

Of course I understand that your interest is to provide a drop-in
solution that works without fixing the clients. It might just not be as
easy is it looks if you rely on existing client software.

Use the open source community to debug your own software, then re-package, but some un-needed useless features into it, and VIOLA ! you have a new proprietary product, thanks to Open Source.

This is why the GNU’s General Public License is nice. Under this license, it is copyright infringement to distribute proprietary software built from GPL’d software without including the source code (which basically eliminates the point of making it proprietary).

To the developers of this software: GPL it, please. Free software is far better than simply “open source” software.

99 Aug 24, 2009 at 19:52 by OneOfThePeople

Brilliant.
I was hoping someone would mention it.

Note, yes this technology is Open Source. But it is a complete and utter sham.
Andreas Dahlström, Peerialism is attempted something quite different here. Peerialism is not interested in torrent, file-sharing, people, community, blahblahblah. No.
Peerialism is simply a new software development model.
Use the open source community to debug your own software, then re-package, but some un-needed useless features into it, and VIOLA ! you have a new proprietary product, thanks to Open Source.

BitTorrent Tracker ??????????????????????????????? ?????????????????????????????????????????? 150% ??……

I’m not interested in what this does to the speed of my torrents, I’m interested in what it could do for the responsiveness of all my other traffic if/when the overseas tubes are decongested.

I can see some major problems with this in regards to improving speeds, as already outlines by others. Aren’t the various ISPs the main reason for slow download speeds?

okay, let’s play a game.
i have a 8mbit/512kbit adsl and i have a limit of ~60KiB/sec upload. using 4 upload slots:
will i have 60KiB/sec upload speed (15KiB/sec/slot) to any part of the world? yes.

now i have a 20mbit/20mbit fiber and 2.5MiB/sec upload limit. using 40 upload slots:
will i have 2.5MiB/sec upload speed (64KiB/sec/slot) to any part of the world? yes. (unless the swarm is extremely healthy and full of seedboxes/fibers)

nothing can help boost bittorrent speeds except for uTP (which will only help on dsls) or getting a better connection.

Andreas: “…what Peerialism gains from it and the simple answer is that we want to show that our technology really can make a difference and also learn from big, real world implementations.”

wow, that’s selfless!
How naive do you think we are, seriously?

you “making a difference” won’t pay the bills, would it? and doesn’t make any sense company-wise!

So how are you planning to pay your employees?

your connection with mil, NATO and the interests involved (as someone else mentioned) would be another point we might be interested in …

There are a only a small handful of ISPs in the world that actually have any benefit at all for locality-based preferencing, and they get it for one of two reasons.
1> Different bandwidth transfer quotas for domestic vs international traffic
2> Different connection speeds for domestic vs international traffic.

In North America, neither of the above two scenarios exist.

I am aware of maybe 15 ISPs worldwide that actually fall into the above two categories that would actually benefit from this.

If more ISPs would cave a little and give us more bandwidth for ISP-Domestic traffic then maybe we would see this have more than just a minor increase on speeds (or, more likely, a major decrease)

In lab conditions, I won’t dispute the possibility of such massive improvements in speeds, but real world conditions are, unfortunately, far enough from lab conditions to not see things work the way you guys are hoping.

Someone asked why we’re doing this or what Peerialism gains from it and the simple answer is that we want to show that our technology really can make a difference and also learn from big, real world implementations.

Regarding the effectiveness of our software we are think the effects will be around the announced numbers but it’s actually hard to know before being tested big scale. This has to our knowledge never been done so far anywhere in the world. We intend to make real life big scale experiments later and evaluate (we’ll keep you posted).

The name “Open Tracker 2.0″ it’s just our internal project name, and it has not been approved by Erdgeist (but we will ask him to). We consider the current Open Tracker the best tracker software in the world – that’s why we wanted it a base for our version. And of course tons of cred goes to Erdgeist :-)

But back to the main question: if we make this new Tracker version and we can prove it really makes Bittorrent better: will the torrent community use it? If not, why? And can we change anything to make it more appealing?

Look, ISPs are between the fronts, since they share interests with both sides.

Rights holders try vehemently to pull them to their side. How will the P2P community prevent this?

We have to make some compromises with possible allies (ISP, content CREATORS) and drive the technology forward :-)

you are already trackable and you have always been. just click on peers tab in utorrent and tah-dah, you “tracked” a lot of people, you nazi!!!!

bittorrent is NOT ANONYMOUS and has NEVER been.

However, I honestly believe that an multi-tracker platform running at each ISP would be the best soultion for both ISPs (a lot less external traffic = reduced costs. BT was responsible for about 50% traffic, right?) and consumers (reduced latency, improved speeds, special plans where you get cheaper traffic rates when P2Ping inside the network). But that is a kinda distant Utopia…

Finally, there are some nice aproximations for finding the best peer other than the IP geolocalization one, such as latency comparation (would need client cooperation) or grouping Peers by ASN (easily done, and pretty effective too). But true location-aware tracking is a lot different…

I do believe that the new Open Tracker will be better, but not “150% faster”

RT
http://www.web-tools.us.tc

Blogpost in the opentracker blog will follow.