North American Network Operators Group

Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical

Re: The large ISPs and Peering

  • From: Majdi S. Abbas
  • Date: Wed Jul 25 18:55:50 2001

On Wed, Jul 25, 2001 at 11:01:57AM -0400, Jeb R. Linton wrote:
> There's nothing sinister or secret about this.

	Except for the obvious troll by the anonymous poster.
If our anonymous poster has the claimed RFPs, why haven't they
been posted?  Why are they posting anonymously?

> I can't say who the winners are because the winners aren't official yet, and
> I also have heard only rumors. The big players are simply doing a smart
> thing - deciding together on points where they can all agree to meet and
> peer at 2.4Gb and 10Gb cheaply. It's obviously the right thing to do.

	I hate to point this out, but the `big players' are invariably 
associated with large LECs, IXCs and other companies in possession of
A) large amounts of fiber  and/or b) large amounts of transport. 

	They are also typically in posession of large amounts of 
colo, usually their own, which they are already deployed into.
Would you rather deploy into your own colo, or someone elses, 
particularly when that someone is in competition with you?  Do you
want to be in a facility your competitors have access to?

> What it means for smaller ISPs, content providers, etc., is that there will
> now be a particular Equinix, Level(3), etc. facility, where we know all the
> big players will be. Those facility providers won't keep us out - they'll
> market the fact that the top Tier-1's are there in order to get everyone
> else there too.

	I think you vastly overestimate the ability of multiple providers
to cooperate on this scale.  The regulatory and goodwill cost of such 
cooperation, even if possible, exceeds the LEC/IXC/fiber subsidized peering 
loops.

> These facilities are huge. Each Tier-1 needs space for a few Juniper M160s,
> Cisco 12400s, etc. The space left is more than enough for Tier-2s and
> content providers galore. There's nothing preventing the big guys from
> competing to provide transit to others in those facilities without huge
> local loop costs. It's basically a one-stop shop for transit circuits from
> anybody you want - they know this, so the competition will be pretty good.

	*bzzt* Wrong again.  In order for a large carrier, particularly 
large access and transit ISPs to deploy into these facilities, they will
require transport equipment.  ADMs and diverse fiber entrances into those
facilities are not cheap.  They would have to fully deploy into each of
these facilities.

	The total equipment, fiber, and facilities cost of such a move is
likely to be much higher than arranging for private peering as needed.

> "What happens to my favorite Co-lo?"... Well, if you're not in the facility
> that gets chosen, it's still likely there will be cheap connections from
> yours to theirs. These thing will sit on multiple metro fiber rings, so
> again there will be decent competition. Any old facility that doesn't hook
> up to the chosen ones knows they will be left out in the dark. So choose
> wisely.

	It comes down to the savings.  From a tier 2 or tier 3 perspective,
transport, particularly large pipes, seems prohibitively expensive.  When
you receive these loops at or below cost due to Ma Telco's subsidization of
those loops, it is very possibly a non-issue to arrange for 'large' circuits,
peering or otherwise, as long as the capacity is there.

	I think your perspective may be skewing your judgement somewhat.

	In my experience, presence at the same facility is only a peering
enhancer for local and regional ISPs, who would very possibly not peer at 
all if there were a loop involved.  The larger carriers have already made
their arrangements.

	--msa