4 ms·
Changing the game with μTP
- jws 17y agoThis new protocol uses packet latency based throttling instead of the packet loss based throttling of TCP. I don't think the article documents any details. Maybe I missed a link.
- Zev 17y agoThe article doesn't mention details but in the comments, Simon links to the IETF working group on this: http://www.ietf.org/dyn/wg/charter/ledbat-charter.html http://www.ietf.org/dyn/wg/charter/ledbat-charter.html
- dschobel 17y agoLatency based congestion control already exists for TCP[1]. So my question is, why wouldn't you get the same effect by running BT on top of a network stack using Vegas? Someone flash the tptacek signal! [1] http://en.wikipedia.org/wiki/TCP_Vegas http://en.wikipedia.org/wiki/TCP_Vegas
- wmf 17y agowhy wouldn't you get the same effect by running BT on top of a network stack using Vegas? You probably would, but Windows doesn't support Vegas. http://www.ietf.org/mail-archive/web/p2pi/current/msg00052.html http://www.ietf.org/mail-archive/web/p2pi/current/msg00052.h...
- jws 17y agoIt appears to me that if you use Vegas on your Linux machine then all of your TCP connections use Vegas, which will make you a second class citizen to all the Reno folk out there on the internet. If you assume that your own uplink is the only bottleneck to be dealt with (say you are serving torrents over a DSL line) then this is reasonable, but I don't think I'd want to yield to all TCP users everywhere with all of my outgoing traffic. There are also some scary comments in the notes from before it was rolled into the kernel, like the code doesn't handle route changes. I can see why a vendor would choose something over which they have control.
- moriarty 17y agoDetails of the protocol are here: http://www.bittorrent.org/beps/bep_0029.html http://www.bittorrent.org/beps/bep_0029.html
- wmf 17y agoWhat's unsaid here is that uTP is an undocumented protocol that isn't available to other BitTorrent implementations like Azureus/Vuze, potentially making them second-class citizens.
- dschobel 17y agofrom the bittorrent.com comments, a quote from a high-up at BT: Simon Morris Posted October 8, 2009 at 5:52 pm As a matter of fact µTP is indeed a replacement for TCP, so we think the name is appropriate. You can check out standardization work which we’re chairing at the IETF called the LEDBAT working group
- wmf 17y agoLEDBAT is only standardizing a small subset of uTP: "Framing is not on the WG's charter." http://www.ietf.org/mail-archive/web/ledbat/current/msg00117.html http://www.ietf.org/mail-archive/web/ledbat/current/msg00117...
- yxhuvud 17y agoThat should be a nonissue as long as the protocol is backwards compatible.
- kidko 17y ago...The way the client talks to other clients has had to be completely re-built. As a side effect, because the new protocol so different, it is practically invisible to some of the nasty traffic shaping techniques that some ISPs have been using. They're not even going to pretend other client will be able to support it in a timely fashion. Backwards compatibility is out the window if they really did implement everything documented in the article.
- sketerpot 17y agoGood news, though: the code is open source.
- 17y ago
- tybris 17y agohttp://research.microsoft.com/en-us/um/people/costa/papers/slot09zeroday.pdf http://research.microsoft.com/en-us/um/people/costa/papers/s... might be an interesting alternative/complement.
- wmf 17y agoOTOH, http://conferences.sigcomm.org/hotnets/2009/papers/hotnets2009-final115.pdf http://conferences.sigcomm.org/hotnets/2009/papers/hotnets20... suggests that localization doesn't help that much.
- known 17y agoIsn't http://en.wikipedia.org/wiki/Precision_Time_Protocol http://en.wikipedia.org/wiki/Precision_Time_Protocol better than uTP?