Leaving public peering?
Henk.Steenman at ams-ix.net
Wed Dec 9 13:41:01 CST 2009
On Dec 3, 2009, at 1:00 AM, Patrick W. Gilmore wrote:
> On Dec 2, 2009, at 4:48 PM, Jonas Frey wrote:
>> the DE-CIX pricing is now 500 Euro/month...since 1st october...see end
>> of that page.
>> Both DE-CIX and AMS-IX have decreased their pricing this year..almost at
>> the same time. I guess this is a move to stop company leaving public
>> exchanges...i have seen this trend, too.
> That is not why LINX lowers its prices. (I cannot say why AMS-IX lowers its prices.)
> LINX is a member-based organization. The member _own_ the exchange. They are paying themselves, and they only pay themselves as much as it costs to run the exchange. With more members, more scale, and advances in equipment, unit (i.e. port) costs go down.
> In a cost-recovery model, that means prices drop.
For exactly the same reason AMS-IX lowered its prices.
> LINX dropped prices mid-year 2009, and are dropping prices again in January 2009. AMS-IX dropped prices once in that time. DE-CIX actually raised its prices for many members, so they could lower their prices for others. Interesting strategy....
>> On Wed, 2009-12-02 at 22:20, Leo Bicknell wrote:
>>> In a message written on Wed, Dec 02, 2009 at 12:46:46PM -0800, Lasher, Donn wrote:
>>>> I realized that paid transit is down at almost obscene levels, but is
>>>> that enough of a reason to increase hop-count, latencies, etc?
>>>> Why disconnect from public mostly-free peering?
>>> Let's look at some economics. I'm going to pick on some folks here,
>>> solely because they have prices online and because they are, I feel,
>>> representative prices.
>>> "Home of the $4 Megabit!" So we have transit prices at $4 per megabit.
>>> A 1GE link to the exchange is 1000 euro per month, which is $1505 USD at
>>> the moment, let's call it $1500 for round numbers.
>>> Now, your 1GE exchange port really shouldn't be run past 60% or so, if
>>> you want to provide good service. So it's really $1500 for 600Mbits,
>>> or $2.50 per Megabit.
>>> If you're an ISP you look at this and go, humm, I take in $4 from my
>>> customer, and hand $2.50 of it right back out to an exchange operator
>>> if I use public peering, making the exchange 62% of my costs right up
>>> front. On the other hand, if I choose wisely where I private peer I
>>> can do it at places with a one-time fee for the cable, so there is
>>> $0 in MRC. I have to buy a router port, sure, but it's also $0 MRC,
>>> just a capital asset that can get written off over many years.
>>> This is the math with the $4 megabit advertised price. The halls at
>>> Nanog are awash in $2 a megabit rumors if you have large enough commits
>>> (say, a few 10GE's). Taking in $2 and paying the exchange operator
>>> $2.50 of it....well, that's not so good. :)
>>> Transit prices have fallen enough that MRC's for switch ports, and
>>> even MRC's for fiber runs (are any of you still in a colo that wants
>>> $500 a month for a fiber run, I didn't think so) are eating up huge
>>> chunks of the inbound revenue, and thus just don't make sense.
>>> Now, before someone points it out, yes, DECIX's rate per megabit is
>>> lower on a 10GE and a second port, so if you can move 2 ports of 10GE of
>>> traffic you can make it a lot cheaper. Also, Cogents $4 a megabit is
>>> probably predicated on you being in the right location and having the
>>> right commit, if you need a DS-3 in West Nowhere you'll pay a higher
>>> rate, and that helps offset some of the costs. I've oversimplified, and
>>> it's a very complex problem for most providers; however I know many are
>>> looking at the fees for peering ports go from being in the noise to a
>>> huge part of their cost structure and that doesn't work.
More information about the NANOG