5 ms·

As a new ISP, you're not going to get any peering arrangements. Really, as an existing consumer ISP not in the top 10 or 20, you're not going to get any peering

As a new ISP, you're not going to get any peering arrangements. Really, as an existing consumer ISP not in the top 10 or 20, you're not going to get any peering arrangements. Consumer ISPs generate very asymmetric traffic (more down than up) which tampers with the core "equality" assumption of traditional peering arrangements. You'll need to purchase transit from an existing provider, such as CenturyLink's commercial arm or Verizon Business, there are plenty more. This won't be free but it also probably won't be as expensive as you think. In a city you can get a Gigabit dedicated to you for $2k a month, maybe less. Shop around. Most of these providers will put you on the phone with a network engineer as part of the quoting process and they can be helpful in recommending what service level and equipment you get. IPv4 allocations are hard to come by unless you want to bid on them at auction which can be expensive. You probably want to get a block from your provider, but they'll be hard to talk into giving you more than say a /29 and they might charge you more for it. Often they don't, though, they just put you in a queue sorted by customer size and you'll never get to the top of it. As a small ISP, CGNAT is probably going to be your only choice for IPv4. This also has the downside of being non-portable since the addresses will always belong to your provider. Maybe save up money to bid on your very own. Once you get one you'll want to start participating in BGP which is complicated anyway, so get to that later, once you have customers.

rahimnathwani 7y ago
'Consumer ISPs generate very asymmetric traffic (more down than up) which tampers with the core "equality" assumption of traditional peering arrangements.' Shouldn't this be in favour of the ISP? If the traffic isn't balanced, then doesn't the peer that's sending more (e.g. YouTube, Netflix) usually pay? I have no experience of setting up an ISP, so this is just based on what I've read. I'd love to understand it better. "IPv4 allocations are hard to come by unless you want to bid on them at auction which can be expensive." Looking online, it looks like the prices are $15 to $30 per IP address, depending on the size of the block. So I guess this cost swallows the first month of revenue from each new customer.
- Scoundreller 7y ago
  > 'Consumer ISPs generate very asymmetric traffic (more down than up) which tampers with the core "equality" assumption of traditional peering arrangements.' It's a good situation for the ISP to start a DC business with unmetered ports to help balance things!
  - jcrawfordor 7y ago
    This is a lot of how the major ISPs smooth out the peering situation, they also offer colocation and carrier hotel services to large customers. Unfortunately, the popularity of Netflix and other streaming services has largely upset this balance and it's very hard for any ISP to keep traffic equality.
    - Scoundreller 7y ago
      At least with Netflix, you can get their Appliances and cut your inbound Netflix bandwidth by 80%+ for the cost of power and 4U.
      - iptrans 7y ago
        Yeah, if you have at least 5 Gbps of Netflix traffic on your network... No small or startup ISP is going to qualify for a Netflix appliance. Also to cut your Netflix bandwidth by 80+%, you need more than 6 Gbps of Netflix traffic, as Netflix requires 1.2 Gbps of inbound bandwidth just to fill the cache. Netflix share of Internet traffic is about 15-20%, so in practice Netflix appliances are used by ISPs with 25-35 Gbps of bandwidth and up. This means ISPs with tens of thousands of subscribers or more.
- jcrawfordor 7y ago
  > Shouldn't this be in favour of the ISP? If the traffic isn't balanced, then doesn't the peer that's sending more (e.g. YouTube, Netflix) usually pay? No, these providers end up peering or placing CDN edge nodes on the provider network and essentially don't pay for their transit. They get this arrangement because it reduces the distance that the high-tier ISPs have to move their traffic. The basic problem is this: ISPs enter peering arrangements with each other to avoid the bureaucracy of both paying each other to access the other's customers. As a consumer ISP, you only have end users, not service providers (or at least you have many more end users than service providers). So, your customers want to access to other networks. No other networks want to access your customers. This means that you don't have the negotiating position to set up peering agreements (because only your own customers would be affected by not having one). In general, peering agreements are becoming very strained over the last five plus years because the underlying assumption of traffic symmetry is virtually never true for consumers now. In practice most of the high-tier ISPs business is providing consumers with access to Netflix. Netflix doesn't care about getting access to anyone else. So no one likes to peer with consumer ISPs, even the big ones, because there's nothing in it for them. This has produced some very high profile disputes with major consumer ISPs dropping peering with tier-1 providers. > Looking online, it looks like the prices are $15 to $30 per IP address, depending on the size of the block. So I guess this cost swallows the first month of revenue from each new customer. Yes, but you don't buy addresses onesy-twosy. If you're going to buy your own block you're probably going to want at least a /24 (255 addresses). That runs upwards of $5k. Most small ISPs start up with very little capital and that's out of reach. Plus you'll need to participate in BGP which is out of the technical expertise of most of the people I know running WISPs, although sometimes you can pay your transit provider to handle BGP for you.

rahimnathwani 7y ago

'Consumer ISPs generate very asymmetric traffic (more down than up) which tampers with the core "equality" assumption of traditional peering arrangements.' Shouldn't this be in favour of the ISP? If the traffic isn't balanced, then doesn't the peer that's sending more (e.g. YouTube, Netflix) usually pay? I have no experience of setting up an ISP, so this is just based on what I've read. I'd love to understand it better. "IPv4 allocations are hard to come by unless you want to bid on them at auction which can be expensive." Looking online, it looks like the prices are $15 to $30 per IP address, depending on the size of the block. So I guess this cost swallows the first month of revenue from each new customer.

Scoundreller 7y ago
> 'Consumer ISPs generate very asymmetric traffic (more down than up) which tampers with the core "equality" assumption of traditional peering arrangements.' It's a good situation for the ISP to start a DC business with unmetered ports to help balance things!
- jcrawfordor 7y ago
  This is a lot of how the major ISPs smooth out the peering situation, they also offer colocation and carrier hotel services to large customers. Unfortunately, the popularity of Netflix and other streaming services has largely upset this balance and it's very hard for any ISP to keep traffic equality.
  - Scoundreller 7y ago
    At least with Netflix, you can get their Appliances and cut your inbound Netflix bandwidth by 80%+ for the cost of power and 4U.
    - iptrans 7y ago
      Yeah, if you have at least 5 Gbps of Netflix traffic on your network... No small or startup ISP is going to qualify for a Netflix appliance. Also to cut your Netflix bandwidth by 80+%, you need more than 6 Gbps of Netflix traffic, as Netflix requires 1.2 Gbps of inbound bandwidth just to fill the cache. Netflix share of Internet traffic is about 15-20%, so in practice Netflix appliances are used by ISPs with 25-35 Gbps of bandwidth and up. This means ISPs with tens of thousands of subscribers or more.
jcrawfordor 7y ago
> Shouldn't this be in favour of the ISP? If the traffic isn't balanced, then doesn't the peer that's sending more (e.g. YouTube, Netflix) usually pay? No, these providers end up peering or placing CDN edge nodes on the provider network and essentially don't pay for their transit. They get this arrangement because it reduces the distance that the high-tier ISPs have to move their traffic. The basic problem is this: ISPs enter peering arrangements with each other to avoid the bureaucracy of both paying each other to access the other's customers. As a consumer ISP, you only have end users, not service providers (or at least you have many more end users than service providers). So, your customers want to access to other networks. No other networks want to access your customers. This means that you don't have the negotiating position to set up peering agreements (because only your own customers would be affected by not having one). In general, peering agreements are becoming very strained over the last five plus years because the underlying assumption of traffic symmetry is virtually never true for consumers now. In practice most of the high-tier ISPs business is providing consumers with access to Netflix. Netflix doesn't care about getting access to anyone else. So no one likes to peer with consumer ISPs, even the big ones, because there's nothing in it for them. This has produced some very high profile disputes with major consumer ISPs dropping peering with tier-1 providers. > Looking online, it looks like the prices are $15 to $30 per IP address, depending on the size of the block. So I guess this cost swallows the first month of revenue from each new customer. Yes, but you don't buy addresses onesy-twosy. If you're going to buy your own block you're probably going to want at least a /24 (255 addresses). That runs upwards of $5k. Most small ISPs start up with very little capital and that's out of reach. Plus you'll need to participate in BGP which is out of the technical expertise of most of the people I know running WISPs, although sometimes you can pay your transit provider to handle BGP for you.