7 ms·
Somebody has to pay the money to upgrade the equipment and bandwidth available at these exchange points. The very reasonable argument in this article is that t
by JoshTriplett 13y ago
Somebody has to pay the money to upgrade the equipment and bandwidth available at these exchange points. The very reasonable argument in this article is that the ISPs should pay that cost, which seems reasonable given that their customers are demanding it. It sounds like the ISPs are playing a game of chicken, trying to see if their peers like Level3 will throw money in to pay for the ISP to upgrade its equipment and bandwidth. That's certainly something the ISPs can try to do; on the other hand, what are their customers going to do, not use Netflix and YouTube? If a pile of customers of one ISP start reporting that they're all having a poor experience with high-bandwidth video, and there are a pile of well-publicized press releases blaming the ISP, customers will start complaining to the ISP, and they'll have to upgrade their infrastructure eventually. (And in areas where they have competition, there's an incentive to upgrade before the competitors, to avoid losing customers; while there isn't such competition in every locale, there are enough locales with more than one ISP choice to make those customers painful to lose.)
But what does any of that have to do with mandated peering requirements at the NSFnet exchanges? Who would enforce that, and why, when any two major networks can set up peering at any number of meet-me rooms? Requiring that an ISP peer as much traffic as is available or not peer any at all seems ridiculous; some ISPs will suck more than others, but that's the problem of them and their customers, not a problem for the entire Internet.
Meanwhile, I'm surprised there aren't more startups and VCs looking to bet that "new ISP that doesn't suck" is a viable business model. People are chomping at the bit for Google Fiber, which seems unlikely to grow to a national level without developing competitors. This is a space with very few competitors, and there hasn't been serious competition in that space since DSL stopped being a viable option.
- bsimpson 13y agoGoogle's ability to print money in conjunction with its generally high PR ratings give it the cache to take on a project like fiber. I can't imagine it's cheap or easy to lay fiber if you don't have substantial financial backing as well as a political goal you're trying to achieve. That said, I've used both Sonic.net and MonkeyBrains here in San Francisco. Never had to rely on a major ISP in this town, and I like it that way.
- CalRobert 13y agoI was unaware these problems were addressed with a large cache. It's a good thing Google has sufficient cachet to persuade public officials to work with them, though!
- diminoten 13y agoThe ISPs aren't dropping the packets, Level3 is. From the article linked to in the blog: "A port that is on average utilised at 90 percent will be saturated, dropping packets, for several hours a day. We have congested ports saturated to those levels with 12 of our 51 peers." "We"in the quote is Level3. They're dropping the packets. Why do the ISPs have to pay for Level3's dropping of packets? Shouldn't the ISPs replace Level3 if they keep dropping packets? Also, dropped packets isn't the end of the world. In TCP at least, the packets just get resent. The language that gets used in all of these talks ("deliberately harming", "American business", "It's insulting") just makes me want to tune out entirely. I don't have time for chicken littles.
- jfoutz 13y agoDropping packets makes load problems worse, because failing users start to send even more traffic.
- MAGZine 13y agoThe opposite, dropped packets are bad because it slows the Internet down. When you need to retransmit a packet because one is dropped, the network assumes congestion and radically throttles speed to prevent the network blowing up.
- Nacraile 13y agoThis is not true. TCP will retransmit, yes, but TCP congestion control will also reduce the transmit rate in response to lost packets. Overall, proper TCP implementations will back off to approximately their fair share of a contended bottleneck. The application layer sees reduced throughput, not message loss. New TCP connections start with very conservative transmit rates, so an application which is overly aggressive in abandoning slow connections and retrying will not add substantially to congestion. UDP doesn't necessarily have these properties, but anyone who implements a high-bandwidth UDP protocol which isn't TCP congestion control friendly deserves public shaming.
- SoftwareMaven 13y ago
- spankalee 13y agoThe problem is that it's easier to compete with Level3 than with Comcast, because while it's incredibly hard to build a backbone provider, it's nearly impossible to build an ISP. This asymmetry is making Comcast bold. They think that eventually competitors to Level3, Netflix, etc., will take advantage of their poor connectivity, pay them, and reap the rewards. They might be right.
- jbooth 13y agoIt could just as easily be the case that nobody's going to put however many dollars into building out a nationwide infrastructure that's completely dependent on future goodwill from Comcast et al, in light of their current behavior. Maybe I'm just an optimist, but I think Level3 wins here just because most people have no idea who they are or what the 'internet backbone' is. "My netflix is slow!" => Netflix's problem. "My internet is slow!" => Cable company's problem.