Internet Exchange Questions

Mailman List Mirror (Read Only)
1

I am a business school student studying the state of the telecom sector and specifically the Internet infrastructure. I am currently trying to understand the role the IX such as PAIX, Equinix, Telehouse, etc… will play in the future where the number of service providers is drastically reduced relative to the environment they were created in. I think PAIX is a good example of this. MFN announced today that they were selling off PAIX. I would be interested in hearing thoughts on why anyone would want to buy PAIX and if there is a way to continue to make money selling cross connects in the future.

Jon Bennett

2

Good question :wink:

You will likely get a lot of different answers to your questions,
depending on who answers, and their experiences in dealing with IX
operators.

My thoughts on the subject are:
1) While the number of service providers (ISPs, MSPs, ASPs, <insert letter
for marketing buzzword>SPs, etc) appears to be on the the decline, it
stands to reason that the amount of traffic present at an exchange point
should remain relatively constant on average, and probably grow. This is
because as service providers go out of business or are acquired by other
organizations, the traffic that a given provider was carrying either gets
displaced by customers taking their traffic/business elsewhere, or the
traffic gets borged into the network of the acquiring provider.

I base this on empirical observation from my viewpoint as a network
engineer at a mid-sized service provider, not on actual observation.
Anyone that has done such observation feel free to chime in :wink:

2) The profitability of an IX is tied to the equipment and design
methodology in use, and of course the amount of traffic provider X can
move on the exchange when they join. Bigger IXs with larger member lists
can equate to better business drivers for providers trying to decide where
and with whom to peer. Also keep in mind that not all providers peer
equally, and some peer more equally than others. Sounds stupid, but it's
true.

Some of the larger, older exchange points, such as the MAEs (not the old
MAEs, mind you), the AADS NAP in Chicago and the PacBell NAP in the Bay
Area are based on ATM. Many newer exchange points are based on Ethernet
(LINX, Equinix, NYIIX/Telehouse, etc). The cost per Ethernet port in
my experience tends to be much cheaper than the cost per ATM port. This
can make setting up an Ethernet-based exchange less costly to set up and
operate.

3) As time passes, more providers either understand the benefits of
peering at an exchange point versus paying ${UPSTREAM} to provide transit
for all of their traffic, or their traffic levels grow to the point (see
point 1) where peering at ${EXCHANGE} begins to make financial sense.
Most providers lack the levels of traffic or the geographic footprint to
peer with the big guys (UUNET, Sprint, AT&T, CW, Genuity, etc), who
typically build private interconnections with each other in multiple
geographically diverse areas. Private interconnects are normally not cost
effective for service providers who don't satisfy those criteria, so for
them, peering at exchange points is more financially/technically
attractive.

As for your question re: PAIX, it is a well-engineered exchange point that
has been around for a long time, with an extensive member list. That,
plus whatever revenue stream PAIX has would probably make an attractive
acquisition for several companies.

The bottom line here is that I think that exchange points will continue to
be technically viable - the Internet traffic that crosses them is substantial,
and not likely to go away any time soon.

hope this helps
jms

3

Once upon a time, Streiner, Justin <streiner@stargate.net> said:

My thoughts on the subject are:
1) While the number of service providers (ISPs, MSPs, ASPs, <insert letter
for marketing buzzword>SPs, etc) appears to be on the the decline, it
stands to reason that the amount of traffic present at an exchange point
should remain relatively constant on average, and probably grow. This is
because as service providers go out of business or are acquired by other
organizations, the traffic that a given provider was carrying either gets
displaced by customers taking their traffic/business elsewhere, or the
traffic gets borged into the network of the acquiring provider.

I would think that, depending on who bought/merged with who, traffic may
decrease as the networks are merged. After all, there was some amount
of traffic between those networks that probably passed through an
exchange point somewhere that is now being handled internally.

I base this on empirical observation from my viewpoint as a network
engineer at a mid-sized service provider, not on actual observation.
Anyone that has done such observation feel free to chime in :wink:

Of course, what do I know; I guess we don't rate as "mid-sized" around
here. :slight_smile:

4

$author = "Streiner, Justin" ;

it stands to reason that the amount of traffic present at an exchange point
should remain relatively constant on average, and probably grow.

someone already mentioned the possibility of traffic that previously
traversed the IX now being interprovider due to an acquisition.

thinking of other ways traffic might decrease, as service providers (of
whatever ilk) fold or are acquired, it might be possible that customers will
shift to one of the big X (where X is the current number of major players aka
"tier 1") either through churn or acquisition.

if that premise holds then traffic that previously may have traversed an IX
between a "tier 2" and a "tier 1" (whether it be peering or transit) might
now be routed via private peering interconnects between two "tier 1"s.

ie. as the market consolidates private interconnects may carry more load at
the expense of IXs.

my au$0.02
marty

5

a message of 72 lines which said:

Most providers lack the levels of traffic or the geographic footprint to
peer with the big guys (UUNET, Sprint, AT&T, CW, Genuity, etc), who
typically build private interconnections with each other in multiple
geographically diverse areas.

A small fix: unlike the other you mention, Genuity accepts peerings
with small providers. You have to do a lot of paper work, you have to
parse legalese, but it works.

6

a message of 21 lines which said:

buy PAIX and if there is a way to continue to make money selling
cross connects in the future.

Not all exchange points are built with the idea of making money. Some
are gratis and some are cheap, with the purpose of improving
connectivity, not making money with the IX. In Paris, France, two IX
(FreeIX and Pouix, may be three with the Sfinx, which is not a
profit-maker) are explicitely built with that aim.

7

When the number of internet exchanges/NAPs/MAEs started to grow in the
early/mid 1990s, many people expected this to be a temporary phenomenon:
surely, broadband ISDN would obsolete them.

This hasn't happened. However, the reasoning still stands: why buy rack
space in a remote place and go through all kinds of trouble to install a
router there, if you can easily use some kind of switched/multiplexed
service from a telco and directly connect with your intended peering
partners over it, regardless of where everyone is located. (Hey, does this
sound like private interconnects?)

This may still happen as ethernet becomes telco-friendlier. But as long as
you're in a location anyway, interconnecting with other networks who are
there as well is always cheaper and easier.

8

This hasn't happened. However, the reasoning still stands: why buy rack
space in a remote place and go through all kinds of trouble to install a
router there, if you can easily use some kind of switched/multiplexed
service from a telco and directly connect with your intended peering

                 ^^^^^

partners over it, regardless of where everyone is located. (Hey, does this
sound like private interconnects?)

You answered your own question. :slight_smile:

9

3) As time passes, more providers either understand
the benefits of
peering at an exchange point versus paying
${UPSTREAM} to provide transit
for all of their traffic, or their traffic levels
grow to the point (see
point 1) where peering at ${EXCHANGE} begins to make
financial sense.
Most providers lack the levels of traffic or the
geographic footprint to
peer with the big guys (UUNET, Sprint, AT&T, CW,
Genuity, etc), who
typically build private interconnections with each
other in multiple
geographically diverse areas. Private interconnects
are normally not cost
effective for service providers who don't satisfy
those criteria, so for
them, peering at exchange points is more
financially/technically
attractive.

Is there a need for additional IXs or are there too
many today and some should be consolidated or shut
down altogether? If there is a need for new IXs, where
do you put them? Who decides where to build a new IX
and how do you get service providers to show up there
once it is built?

Thanks.

10

> 3) As time passes, more providers either understand
> the benefits of
> peering at an exchange point versus paying
> ${UPSTREAM} to provide transit
> for all of their traffic, or their traffic levels
> grow to the point (see
> point 1) where peering at ${EXCHANGE} begins to make
> financial sense.
> Most providers lack the levels of traffic or the
> geographic footprint to
> peer with the big guys (UUNET, Sprint, AT&T, CW,
> Genuity, etc), who
> typically build private interconnections with each
> other in multiple
> geographically diverse areas. Private interconnects
> are normally not cost
> effective for service providers who don't satisfy
> those criteria, so for
> them, peering at exchange points is more
> financially/technically
> attractive.
>

Is there a need for additional IXs or are there too many today and some
should be consolidated or shut down altogether?

For clarification I may use the terms IX and exchange point
interchangeably.

That would depend on each exchange point operator. There are now
many smaller regional exchange points in areas that are far enough away
from the 'big' exchange points that are likely to be self sustaining
because of local interest. Many of them will probably not grow bigger
than that, but I'd hazard a guess that the operators of the smaller
exchanges didn't set out to become the next MAE-EAST (ok, bad example :wink: )
but rather to improve interconnectivity between local/regional companies
for the mutual benefit of all exchange members, often on a cost-recovery
basis.

http://www.ep.net/ lists many exchange points around the world, large and
small.

If there is a need for new IXs, where do you put them? Who decides
where to build a new IX and how do you get service providers to show up
there once it is built?

These days, that can be a chicken-and-egg question. There really isn't a
formalized process for deciding where an IX should be placed. In the case
of some of the regional points, they came about because someone took the
initiative to build them.

I'd imagine if you're located in a city where:
1) The cost of a circuit to the nearest exchange point is too high
2) There are a decent number of local organizations who may be interested
  in or capable of peering there

then it may make sense to establish an IX. It doesn't take much to run a
small one. You could do it with an Ethernet switch and a small amount of
power and rack space to start.

It may also help if you're not a service provider yourself. Sometimes
local providers get standoffish about peering at an IX run by a
competitor. Strange, but sometimes so is human/social psychology :wink:

If can also help if your area/building is served by more than one telco.

Getting providers to show up usually starts with some type of grass-roots
effort, getting the word out on the street that you have a place where
local providers can meet to exchange traffic. Getting to the point there
the IX assumes a critical mass of sorts and makes sense to operate for the
long term takes time and effort.

jms

11

That's certainly a possibility. I was thinking about the traffic
displacement on top of regular growth.

jms

12

the "if you build it they will come.." strategy has
surely passed. we will likely see some continued
consolidation as the telcos restructure (ie paix to be
sold yet again.) capacity and usage within the
ramaining ixps will dictate the need for future growth.

  /rf

13

And for European IXPs there is now an association:

  http://www.euro-ix.net/

Which has details of the member exchanges.

  f

14

Hi,

    I don't know if the question was mainly intended towards the american
based IXes. If so, then the below may be of little interest. If not,
please read on.

15

As for your question re: PAIX, it is a well-engineered exchange point that
has been around for a long time, with an extensive member list. That,
plus whatever revenue stream PAIX has would probably make an attractive
acquisition for several companies.

Thank you for those kind words. And if you've ever got a problem with PAIX,
you know who to yell at.

16

Is there a need for additional IXs or are there too
many today and some should be consolidated or shut
down altogether? If there is a need for new IXs, where
do you put them? Who decides where to build a new IX
and how do you get service providers to show up there
once it is built?

Speaking for PAIX, we build where our customers tell us to.

17

Is there a need for additional IXs or are there too

    > many today and some should be consolidated or shut
    > down altogether? If there is a need for new IXs, where
    > do you put them? Who decides where to build a new IX
    > and how do you get service providers to show up there
    > once it is built?

I know of 323 IXes today.
http://www.pch.net/documents/data/exchange-points/ep-in-addrs.xls

In my experience, there are a few states that local exchange "markets" go
through:

First, no exchange, or something which claims to be one but isn't.

    Then, inspiration strikes someone, and:

Second, there's an exchange, and life is good.

    Then, idiocy strikes someone else, they determine that the existence
    of an exchange "validates the market" for exchange services in that
    region, and:

Third, multiple exchanges spring up in the same area, tearing apart the
switch fabric and ruining the economic reason for peering in the first
place.

    Then, people grow to understand the market, and:

Fourth, the different exchange-like services differentiate sufficiently
that people can use peering-oriented exchanges principally for peering,
and transit-oriented exchanges principally for transit.

                                -Bill

18

Until MFN sells them in coming months in their attempts to pay off
billions of dollars of debt...

This industry is so far in the shitter...so many of the big names,
including players from the early days, are in chapter 11 or about to be.

It's a sad day when Qwest looks like a good company.

Andy

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Andy Dills 301-682-9972
Xecunet, LLC www.xecu.net
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Dialup * Webhosting * E-Commerce * High-Speed Access

19

I'm honestly surprised that I haven't had someone try to offer me a
'genuine steal' on 20-year IRUs in awhile :wink:

jms

20

> you know who to yell at.

Until MFN sells them in coming months in their attempts to pay off
billions of dollars of debt...

No change is expected in "who you yell at if PAIX isn't doing a good job."

(That is, "me.")