Comcast Discloses Throttling Practices 206
Wired reports that Comcast finally provided information on its network management practices late Friday. In a report to the FCC (PDF), the cable company admitted to targeting P2P protocols Ares, BitTorrent, eDonkey, FasTrack, and Gnutella. Quoting:
"For each of the managed P2P protocols, the [Sandvine Policy Traffic Switch] monitors and identifies the number of simultaneous unidirectional uploads that are passed from the [Cable Modem Termination System] to the upstream router. Because of the prevalence of P2P traffic on the upstream portion of our network, the number of simultaneous unidirectional upload sessions of any particular P2P protocol at any given time serves as a useful proxy for determining the level of overall network congestion. For each of the protocols, a session threshold is in place that is intended to provide for equivalently fair access between the protocols, but still mitigate the likelihood of congestion that could cause service degradation for our customers."
Now what will happen? (Score:1, Informative)
Comcast will enforce bandwidth caps.
How's that better than throttling?
Re:I have a sneaking suspicion (Score:2, Informative)
No.
Its more like they're saying "traffic outside our network will cost you; internal stuff is free".
In other words it is no different to the way many ISPs behave in the UK. They have mirrors of things people might want to use - so that their customers don't use more external bandwidth than they need to.
For example Virgin Media's Debian mirror [virginmedia.com].
Compare to AU/NZ cap policies (Score:3, Informative)
That the 250GB limit will not be applied to traffic within Comcast's own network. Can you say anticompetitive?
As I understand it, it's fairly commonplace in the Internet access industry not to charge end users for traffic that doesn't cross the ISP's upstream connection. For example, ISPs in Australia and New Zealand, two countries that have a slow, expensive pipe to other anglophone countries (USA, Canada, Ireland, UK), follow this policy of not counting accesses to, say, Linux distro mirrors on the ISP's network against the user's cap.
Re:interpreted to mean (Score:2, Informative)
Re:Almost Worse than Legalese (Score:5, Informative)
"For each of the managed P2P protocols, the [Sandvine Policy Traffic Switch] monitors and identifies the number of simultaneous unidirectional uploads that are passed from the [Cable Modem Termination System] to the upstream router.
Sandvine checks uploads without downloads. It does this 'above' (in the hierarchy) from the head-end of the cable network (neighborhood box).
Because of the prevalence of P2P traffic on the upstream portion of our network, the number of simultaneous unidirectional upload sessions of any particular P2P protocol at any given time serves as a useful proxy for determining the level of overall network congestion.
P2P is used a lot, and fairly consistently. Therefore, the number of one-way uploads (not SSH or rdesktop like somebody else said) can be used to extrapolate the total congestion for much less 'thought' (for Sandvine)
For each of the protocols, a session threshold is in place that is intended to provide for equivalently fair access between the protocols, but still mitigate the likelihood of congestion that could cause service degradation for our customers."
We count the number (like, only 500 BitTorrent sessions) and cut off after that.
--
My thoughts: I don't think this helps anything. I doubt anybody has much of a problem with them legitimately throttling P2P protocols, as long as it's done consistently and fairly (no need to throttle with plenty of upstream, right?). The real problem are the RSTs which impersonate each side of the connection to the other, saying that the other closed the connection. That's like Bob passing messages between Alice and Candice, and telling Candice that Alice called her a bitch, and telling Alice that Candice called her a bitch.
QoS isn't that hard, and I'm sure they know how. It's fairly easy to throttle back without sending RSTs, and allows for the full utilization of 'open' bandwidth.
This statement explains the rationale, but not the choice of methods.
Re:Evil from cable companies? Nevar. (Score:5, Informative)
Comcast hasn't advertised "unlimited internet" in many years. After a Google search, the only use of "unlimited" I could find in a current Comcast ad was associated with their phone service: "Make unlimited local and long distance calls with 12 popular features..."
Re:Now what will happen? (Score:5, Informative)
How is this announcement related to the recent 250 GB monthly usage threshold?
The two are completely separate and distinct. The new congestion management technique is based on real-time Internet activity. The goal is to avoid congestion on our network that is being caused by the heaviest users. The technique is different from the recent announcement that 250 GB/month is the aggregate monthly usage threshold that defines excessive use.
Gizmodo's take on the thing [gizmodo.com] is much easier to read.
Going over the 250GB cap will get you disconnected, but your bandwidth will get throttled long, long before that if you do anything their software deems "excessive."
Comcast problem in Denver (Score:2, Informative)
Re:Evil from cable companies? Nevar. (Score:5, Informative)
The thing is, time is no longer an issue in modern connections because they are packet-switched down to the bare wire.
In the old days you used a phone line, which was circuit switched, to call your ISP. They had a limited number of ports so they had to limit how long you could be online, otherwise folks would get a busy signal.
Since these days there is no customer-initiated circuit switching involved in cable and DSL links, the concept of "unlimited" can *only* apply to data transfer. There isn't anything else to limit.
Believe me, I remember the days of circuit switching and "hourly limits" quite well. I was on an ISDN connection from 2000 to 2004. Worrying about how *long* you're online is extremely irritating. Those are definitely "good old days" I wouldn't want to go back to.
defeating sandvine with packet filter software (Score:1, Informative)
There are a few sites like this now, but I hope this can work for people
http://www.overclock.net/networking-security/276902-sandvine-fix.html
Re:They don't Throttle, they Forge Reset Packets (Score:3, Informative)
That's Comcast old (current) policy. Their new policy is documented in this page [comcast.net] on their web site.
On page 11:
"As described above, the new approach will not manage congestion by focusing on managing the use of specific protocols. Nor will this approach use 'reset packets.'"
what about when speakeasy lies to you? (Score:4, Informative)
There ARE people lying out there. Plenty.
Re:Take it, leave it, or leave it (Score:3, Informative)
Nice talking point. However, the current economic crisis has nothing to do with mandated lending. It's entirely caused by voluntary risks and outright fraud.