A bunch of us are thinking about multihoming solutions for IPv6. For this purpose, it is useful to know a bit more about how actual networks (rather than the ones existing only as ASCII drawings) interconnect. So: - What are the 12 - 18 most important interconnect locations in the world? MAE East, the Ameritech, Sprint and PacBell NAPs, PAIX, LINX and AMS-IX come to mind, but from where I'm sitting it's hard to judge whether others are important or marginal. - To how many of them do typical tier-1 and tier-2 networks connect? - Using private or public interconnects?
Hi Iljitsch, I would not consider Sprint NAP, a place closed to new customers for several years, an important interconnect location in the US. ATM based IXs are not as participant rich as they were 2-3 years ago. The fastest growing US interconnect locations are cross-connect enabled. PAIX & Equinix. Equinix-Ashburn, PAIX-Seattle, Equinix-Newark and Equinix-Dallas and others have seen participation grow with a diverse blend of traffic from cable operators, telcos and content providers. Tier-1 means what? Look for growing sources of traffic. Your mileage may vary, -ren At 11:48 AM 5/17/2002 +0200, Iljitsch van Beijnum wrote:
A bunch of us are thinking about multihoming solutions for IPv6. For this purpose, it is useful to know a bit more about how actual networks (rather than the ones existing only as ASCII drawings) interconnect. So:
- What are the 12 - 18 most important interconnect locations in the world? MAE East, the Ameritech, Sprint and PacBell NAPs, PAIX, LINX and AMS-IX come to mind, but from where I'm sitting it's hard to judge whether others are important or marginal.
- To how many of them do typical tier-1 and tier-2 networks connect?
- Using private or public interconnects?
What about NYIIX/6IIX? Being in Telehouse where there are no monthly fees for for cross-connects gives it a financial advantage over Equinix. Ralph Doncaster principal, IStop.com div. of Doncaster Consulting Inc. On Fri, 17 May 2002, ren wrote:
Hi Iljitsch,
I would not consider Sprint NAP, a place closed to new customers for several years, an important interconnect location in the US. ATM based IXs are not as participant rich as they were 2-3 years ago.
The fastest growing US interconnect locations are cross-connect enabled. PAIX & Equinix. Equinix-Ashburn, PAIX-Seattle, Equinix-Newark and Equinix-Dallas and others have seen participation grow with a diverse blend of traffic from cable operators, telcos and content providers.
Tier-1 means what? Look for growing sources of traffic.
Your mileage may vary, -ren
At 11:48 AM 5/17/2002 +0200, Iljitsch van Beijnum wrote:
A bunch of us are thinking about multihoming solutions for IPv6. For this purpose, it is useful to know a bit more about how actual networks (rather than the ones existing only as ASCII drawings) interconnect. So:
- What are the 12 - 18 most important interconnect locations in the world? MAE East, the Ameritech, Sprint and PacBell NAPs, PAIX, LINX and AMS-IX come to mind, but from where I'm sitting it's hard to judge whether others are important or marginal.
- To how many of them do typical tier-1 and tier-2 networks connect?
- Using private or public interconnects?
On Fri, 17 May 2002, Ralph Doncaster wrote:
What about NYIIX/6IIX? Being in Telehouse where there are no monthly fees for for cross-connects gives it a financial advantage over Equinix.
While I agree, IIX relatively speaking is small -- aggregating about 450 to 500 mb/s. Also, you don't find many US-based internation networks there (ie, UU, Sprint, CW, PSI/Cogent, etc.); however, the participation of Asian and European networks is very impressive. Also, the IIX is run the way I like a NAP run (as if my opinion matters on this); cheap, simplistic, and reliable. I don't know of any other NAP that can claim all three.
Ralph Doncaster principal, IStop.com div. of Doncaster Consulting Inc.
On Fri, 17 May 2002, ren wrote:
Hi Iljitsch,
I would not consider Sprint NAP, a place closed to new customers for several years, an important interconnect location in the US. ATM based IXs are not as participant rich as they were 2-3 years ago.
The fastest growing US interconnect locations are cross-connect enabled. PAIX & Equinix. Equinix-Ashburn, PAIX-Seattle, Equinix-Newark and Equinix-Dallas and others have seen participation grow with a diverse blend of traffic from cable operators, telcos and content providers.
Tier-1 means what? Look for growing sources of traffic.
Your mileage may vary, -ren
At 11:48 AM 5/17/2002 +0200, Iljitsch van Beijnum wrote:
A bunch of us are thinking about multihoming solutions for IPv6. For this purpose, it is useful to know a bit more about how actual networks (rather than the ones existing only as ASCII drawings) interconnect. So:
- What are the 12 - 18 most important interconnect locations in the world? MAE East, the Ameritech, Sprint and PacBell NAPs, PAIX, LINX and AMS-IX come to mind, but from where I'm sitting it's hard to judge whether others are important or marginal.
- To how many of them do typical tier-1 and tier-2 networks connect?
- Using private or public interconnects?
-- Alex Rubenstein, AR97, K2AHR, alex@nac.net, latency, Al Reuben -- -- Net Access Corporation, 800-NET-ME-36, http://www.nac.net --
On Fri, 17 May 2002, Alex Rubenstein wrote:
On Fri, 17 May 2002, Ralph Doncaster wrote:
What about NYIIX/6IIX? Being in Telehouse where there are no monthly fees for for cross-connects gives it a financial advantage over Equinix.
While I agree, IIX relatively speaking is small -- aggregating about 450 to 500 mb/s.
Statistics for the NYIIX switch are very misleading due to the fact that Telehouse NY has such cost effective cross connect fees. Most of the networks there have several private interconnects to their other peers there. Due to the economics compared to exchanges with a similar number of participants and higher cross connect fees, each peer at NYIIX is likely to be more predisposed to establishing private interconnects with their largest traffic flow peers. Telehouse is a great facility to deal with. Also, when considering where to locate equipment I can't help but remind the other person posting about MNFX (which isn't the same corporate entitiy as PAIX) that they should have posted (if they were going to bother to post that kind of thing, which doesn't make friends ;) similar press releases regarding their other choices, or perhaps considering whether the companies they consider alternatives are EBITDA postive (making a profit, or in otherwords will exist in 12 months) today (not in an imaginary planned future) or for the few that are EBITDA positive, whether they actually seem to want your business. PAIX, Palo Alto is another excellent exchange point. I'm sure you can probably get a rack now with all the bankruptcies. If not send me email. Hurricane has IPv6 routers deployed nationally, we are participating in the IPv6 exchange at PAIX, are connected to 6TAP at AADS, and have an IPv6 router at Telehouse. We will be connecting to the 6IIX/NY6IX next week. We are happy to peer natively at any facility we are at or over IPv4 tunnels. We are also happy to allocate a /48 to anybody that is working on getting their pTLA (6BONE) or subTLA (ARIN) (You have to be operational and deployed in order to request and receive your own IPv6 address space from ARIN or 6BONE.) Mike. +------------------- H U R R I C A N E - E L E C T R I C -------------------+ | Mike Leber Direct Internet Connections Voice 510 580 4100 | | Hurricane Electric Web Hosting Colocation Fax 510 580 4151 | | mleber@he.net http://www.he.net | +---------------------------------------------------------------------------+
On Sat, 18 May 2002, Mike Leber wrote:
press releases regarding their other choices, or perhaps considering whether the companies they consider alternatives are EBITDA postive (making a profit, or in otherwords will exist in 12 months) today (not in an imaginary planned future) or for the few that are EBITDA positive, whether they actually seem to want your business.
"EBITDA positive" does not mean profitable, or even necessarily financially stable. EBITDA is earnings before interest, taxes, depreciation, and amoritization -- all things that tend to have an impact on your finances. If you were using EBITDA as the measure of your personal financial situation, you could spend far more than your after tax income, but less than your before tax income, and declare yourself to have come out ahead. Your bank, however, probably wouldn't see it that way. The same goes for corporate finance, except that the corporations that were announcing their EBITDA numbers as the important financial data often had enough in the bank, and enough market cap, that it didn't become a critical problem for a few years. My understanding is that EBITDA does have legitimate accounting uses, but I'm not clear on what they are. I'm tempted to label this message as off-topic nitpicking, but given that the biggest problem with Internet stability at the moment seems to be financial, I'm not sure it is. -Steve -------------------------------------------------------------------------------- Steve Gibbard scg@gibbard.org
On Sat, 18 May 2002, Steve Gibbard wrote:
"EBITDA positive" does not mean profitable, or even necessarily financially stable. EBITDA is earnings before interest, taxes, depreciation, and amoritization
Correct, however I was trying to provide a simplified translation. A company that isn't EBITDA positive can't survive by declaring bankruptcy becausee even after they get rid of the interest payments they will still have a negative run rate. The reason for using EBITDA as an early indicator for financial health when analyzing companies is that it allows you to look at the health of the operation independent of their debt structure and prior capital expenditures (depreciation and amortization) so that you can get a better idea of their cash flow. The reason why cash flow matters is because when a company runs out of cash bankruptcy is imminent. Profitiability from a P&L statement (expecially for public companies) involves so many components that it frequently doesn't allow you to evaluate a company until it has matured.
The same goes for corporate finance, except that the corporations that were announcing their EBITDA numbers as the important financial data often had enough in the bank, and enough market cap, that it didn't become a critical problem for a few years.
True, however by looking at EBITDA and current assets (cash in the bank) you can get a quick picture of the likely hood a company solving anything by declaring bankruptcy and a rough time frame to their imminent demise.
My understanding is that EBITDA does have legitimate accounting uses, but I'm not clear on what they are.
I hope you find my explanation above a useful rule of thumb.
I'm tempted to label this message as off-topic nitpicking, but given that the biggest problem with Internet stability at the moment seems to be financial, I'm not sure it is.
Due to the fact that I've had to order redundant capacity from multiple vendors in situations where there was enough traditional physical network redundancy, this seems to have become an important network provisioning issue. Mike. +------------------- H U R R I C A N E - E L E C T R I C -------------------+ | Mike Leber Direct Internet Connections Voice 510 580 4100 | | Hurricane Electric Web Hosting Colocation Fax 510 580 4151 | | mleber@he.net http://www.he.net | +---------------------------------------------------------------------------+
"EBITDA positive" does not mean profitable, or even necessarily financially stable.
Right you are. So please let me clarify my earlier statement (that "PAIX has been modestly profitable for years"). If we were not a wholly owned subsidiary we would owe income taxes. When we have been wholly owned by companies who were paying income taxes, some of the taxes they had to pay were because of PAIX. (Presumably this positive our income situation will make it easy for MFN to sell us.) Let's have a look at Extreme Networks' recently published financials. (Bring up http://biz.yahoo.com/fin/l/e/extr.html to follow along.) These folks showed a net loss this quarter yet the analysts applauded them and their stock shot up a bit because they had a nonrecurring charge larger than their net loss. This tells analysts that the company would have taxable income if not for the nonrecurring event, which gives them hope for the next quarter. On http://biz.yahoo.com/fin/l/e/extr_ai.html we even see that in the year ending July 2000 they paid $10M in income taxes, which tells us that maybe they know how that feels and want to do it again some day. I like EBITDA as a yardstick for measuring one company against another if they are otherwise similar and I'm looking for a differentiator. But I don't personally buy stock based on EBITDA numbers -- I want to see actual net income and, paradoxically, I love a company who has to pay income tax because it means they had INCOME to pay taxes on. -- Paul Vixie <vixie@eng.paix.net> President, PAIX.Net Inc. (NASD:MFNX)
The main fallacy of EBITDA is that a lot of people confuse EBIDTA figures with cash flow figures. While the utility of a quarterly figure showing cash flow P&L, stripping off all noncash transactions, would be substantial, most companies prefer to quote EBIDTA instead, which, while disregarding all noncash figures, also removes interest and taxes as well, both of which are very much recurring cash expenditures and should be included in cash-flow P&L figures. In the absence of a cash-flow P/L figure, a lot of people look at EBITDA instead and forget about the very real cash expenditures involved with interest and taxes (and often other case expenditures that the company chooses to throw out in order to make the number look better). Intermedia, for example, was EBITDA positive for all of the time I was working for them, yet was bleeding approx. $100 million plus in interest payments per year. This created a very real cash crunch that prompted the sale to Worldcom. -C On Sat, May 18, 2002 at 06:09:56PM -0700, Steve Gibbard wrote:
On Sat, 18 May 2002, Mike Leber wrote:
press releases regarding their other choices, or perhaps considering whether the companies they consider alternatives are EBITDA postive (making a profit, or in otherwords will exist in 12 months) today (not in an imaginary planned future) or for the few that are EBITDA positive, whether they actually seem to want your business.
"EBITDA positive" does not mean profitable, or even necessarily financially stable. EBITDA is earnings before interest, taxes, depreciation, and amoritization -- all things that tend to have an impact on your finances. If you were using EBITDA as the measure of your personal financial situation, you could spend far more than your after tax income, but less than your before tax income, and declare yourself to have come out ahead. Your bank, however, probably wouldn't see it that way. The same goes for corporate finance, except that the corporations that were announcing their EBITDA numbers as the important financial data often had enough in the bank, and enough market cap, that it didn't become a critical problem for a few years.
My understanding is that EBITDA does have legitimate accounting uses, but I'm not clear on what they are.
I'm tempted to label this message as off-topic nitpicking, but given that the biggest problem with Internet stability at the moment seems to be financial, I'm not sure it is.
-Steve
-------------------------------------------------------------------------------- Steve Gibbard scg@gibbard.org
On Mon, 20 May 2002 12:08:32 EDT, Chris Woodfield said:
Intermedia, for example, was EBITDA positive for all of the time I was working for them, yet was bleeding approx. $100 million plus in interest payments per year. This created a very real cash crunch that prompted the sale to Worldcom.
I believe the *original* comment was "If they're EBITDA-negative, they're *really* screwed without more cash(*)". -- Valdis Kletnieks Computer Systems Senior Engineer Virginia Tech (*) As many dot-bombed discovered when the bubble burst...
My take on ebitda, it is what non profitable companies use to put a positive spin on their situation. Bri On Mon, 20 May 2002, Chris Woodfield wrote:
The main fallacy of EBITDA is that a lot of people confuse EBIDTA figures with cash flow figures. While the utility of a quarterly figure showing cash flow P&L, stripping off all noncash transactions, would be substantial, most companies prefer to quote EBIDTA instead, which, while disregarding all noncash figures, also removes interest and taxes as well, both of which are very much recurring cash expenditures and should be included in cash-flow P&L figures. In the absence of a cash-flow P/L figure, a lot of people look at EBITDA instead and forget about the very real cash expenditures involved with interest and taxes (and often other case expenditures that the company chooses to throw out in order to make the number look better).
Intermedia, for example, was EBITDA positive for all of the time I was working for them, yet was bleeding approx. $100 million plus in interest payments per year. This created a very real cash crunch that prompted the sale to Worldcom.
-C
On Sat, May 18, 2002 at 06:09:56PM -0700, Steve Gibbard wrote:
On Sat, 18 May 2002, Mike Leber wrote:
press releases regarding their other choices, or perhaps considering whether the companies they consider alternatives are EBITDA postive (making a profit, or in otherwords will exist in 12 months) today (not in an imaginary planned future) or for the few that are EBITDA positive, whether they actually seem to want your business.
"EBITDA positive" does not mean profitable, or even necessarily financially stable. EBITDA is earnings before interest, taxes, depreciation, and amoritization -- all things that tend to have an impact on your finances. If you were using EBITDA as the measure of your personal financial situation, you could spend far more than your after tax income, but less than your before tax income, and declare yourself to have come out ahead. Your bank, however, probably wouldn't see it that way. The same goes for corporate finance, except that the corporations that were announcing their EBITDA numbers as the important financial data often had enough in the bank, and enough market cap, that it didn't become a critical problem for a few years.
My understanding is that EBITDA does have legitimate accounting uses, but I'm not clear on what they are.
I'm tempted to label this message as off-topic nitpicking, but given that the biggest problem with Internet stability at the moment seems to be financial, I'm not sure it is.
-Steve
-------------------------------------------------------------------------------- Steve Gibbard scg@gibbard.org
On Sat, 18 May 2002, Mike Leber wrote:
Also, when considering where to locate equipment I can't help but remind the other person posting about MNFX (which isn't the same corporate entitiy as PAIX) that they should have posted (if they were going to bother to post that kind of thing, which doesn't make friends ;) similar press releases regarding their other choices, or perhaps considering whether the companies they consider alternatives are EBITDA postive (making a profit, or in otherwords will exist in 12 months) today (not in an imaginary planned future) or for the few that are EBITDA positive, whether they actually seem to want your business.
PAIX, Palo Alto is another excellent exchange point. I'm sure you can probably get a rack now with all the bankruptcies. If not send me email.
At present, PAIX is owned by MFNX, and so is Abovenet. Arms-length or not, ultimately it's the same ailing giant with clay feet... The "other person posting" is glad to oblige by reposting a usenet news article written by an eyewitness who visited Metromedia Fiber/Abovenet's San Jose II co-lo facility. Consider it friendly fire from some anti-spam activist (credible source in your camp, if we were to believe the popular sentiment?):
From: spambait@petra.dyndns.org (Cameron L. Spitzer) Newsgroups: news.admin.net-abuse.email Subject: Re: Abovenet/MFN dead at the switch Date: 23 Mar 2002 19:11:33 GMT Organization: http://extra.newsguy.com Lines: 30 Message-ID: <slrna9pkr6.1eo.spambait@pk.greens.org> References: <8935cf7e.0203230809.30d2bfd1@posting.google.com> NNTP-Posting-Host: p-884.newsdawg.com X-Warning: I take time to damage spammers. User-Agent: slrn/0.9.6.2 (Linux)
In article <8935cf7e.0203230809.30d2bfd1@posting.google.com>, No spammage wrote:
Are there any humans at Above.net/MFN, or is it all spambotically controlled? I keep getting DirecTV spam for 208.185.127.162 (ofertadirectv.iwarp.com) and the ignorebot keeps telling me that Above.net is happy with the way it's been settled. Are they completely black hat?
I installed a server at their San Jose II co-lo in summer of 2000. It was a beehive, with full cages and people carting stuff in and lots of activity. The area where my ISP's rack was was nearly full. My ISP switched providers a year later, because Above's senior technical staff who had attracted them there had all left and the service had deteriorated badly. Almost all of the equipment in the cages around us had already been removed. The floor was really dirty and there was trash everywhere, it looked like the flea market when most of the vendors have left for the day. While I was waiting for my ISP to come and badge me in, I talked to the security guard. He said they'd recently laid off 30% and expected more. He also said he hadn't seen any customer bring any equipment into that facility in the last three months, but he saw people removing stuff daily. They're not black hat, they're empty hat. There's nobody left, and the last guy forgot to turn out the lights. Cameron
(sorry if you see it twice)
A bunch of us are thinking about multihoming solutions for IPv6. For this purpose, it is useful to know a bit more about how actual networks (rather than the ones existing only as ASCII drawings) interconnect. So:
- What are the 12 - 18 most important interconnect locations in the world? MAE East, the Ameritech, Sprint and PacBell NAPs, PAIX, LINX and AMS-IX come to mind, but from where I'm sitting it's hard to judge whether others are important or marginal.
NSPIXP6 (tokyo) has the largest number of participants (40+). 6TAP, NY6IX, 6IIX-NY, and PAIX palo alto comes to my mind next. check out http://www.v6nap.net/.
- Using private or public interconnects?
both. there are a lot of private interconnections over tunnel, which should be changed to CAT5 cable. itojun
Tier-1 means what?
Lately, 'Tier-1' and '[near] bankruptcy' seem to be interchangable. -- Alex Rubenstein, AR97, K2AHR, alex@nac.net, latency, Al Reuben -- -- Net Access Corporation, 800-NET-ME-36, http://www.nac.net --
If that is true then everybody is a Tier-1 carrier. alex@nac.net wrote:
Tier-1 means what?
Lately, 'Tier-1' and '[near] bankruptcy' seem to be interchangable.
-- Alex Rubenstein, AR97, K2AHR, alex@nac.net, latency, Al Reuben -- -- Net Access Corporation, 800-NET-ME-36, http://www.nac.net --
ADC> Date: Fri, 17 May 2002 10:50:29 -0400 ADC> From: Anthony D Cennami ADC> If that is true then everybody is a Tier-1 carrier. Well, it seems that most everybody claims to be "Tier 1". Maybe the bad karma is coming back to bite. ;-) -- Eddy Brotsman & Dreger, Inc. - EverQuick Internet Division Phone: +1 (316) 794-8922 Wichita/(Inter)national Phone: +1 (785) 865-5885 Lawrence ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Date: Mon, 21 May 2001 11:23:58 +0000 (GMT) From: A Trap <blacklist@brics.com> To: blacklist@brics.com Subject: Please ignore this portion of my mail signature. These last few lines are a trap for address-harvesting spambots. Do NOT send mail to <blacklist@brics.com>, or you are likely to be blocked.
perhaps better late than never... PAIX & LINX both have IPv6 capabilities at/on the exchange fabric(s). I am not aware that Equinix has taken that step. --bill
On Fri, 17 May 2002 bmanning@karoshi.com wrote:
perhaps better late than never... PAIX & LINX both have IPv6 capabilities at/on the exchange fabric(s). I am not aware that Equinix has taken that step.
Uhm, another dumb question. Why does the operator of a layer 2 exchange care (or know) what protocols your are using? IPv4, IPv6, heck I remember seeing Appletalk, OSI and DECNET on MAE-EAST. What consenting network operators do.... What step does Equinix (or any other layer 2 exchange) need to do? The ATM NAPs might have an issue due to ATM/ARP, but even then I suspect two consenting network operators could use static IPv6 ARP tables without the NAP operator doing anything.
Why does the operator of a layer 2 exchange care (or know) what protocols your are using? IPv4, IPv6, heck I remember seeing Appletalk, OSI and DECNET on MAE-EAST. What consenting network operators do....
What consenting network operators do bilaterally in an L2 environment where their actions might possibly affect other customers of the L2 exchange is of great interest to the exchange point provider. Take multicast at an exchange point, for instance. IGMP is an edge protocol, not used in the exchange of multicast information between autonomous systems. Since there's no IGMP to snoop, multicast packets are flooded to all ports of an Ethernet-based L2 fabric. Combine that fact with the fact that a range of port speeds are offered, and stir: you could get into a sitation where multicast traffic solicited by one high-speed port customer of another high-speed port customer "drowns out" lower-speed port customers (especially when available speeds span two orders of magnitude). For this reason, multicast traffic exchange on Ethernet-based L2 exchange points is often conducted on a separate switch fabric, or a separate VLAN (as in the case of PAIX). When we see PIM on the main "unicast" VLAN, we ask the PIM speaker to take it to the multicast VLAN (for which we provide IPv4 address assignment just like the other). "Consenting network operators" also engage in practices such as connecting their exchange point switch ports into an aggregation switch of their own, and then connecting their aggregation switches together to implement private peering. If not properly configured and maintained, this has the potential to introduce loops in the switching fabric that can lead to a variety of failure modes for other participants. Engineering an L2 fabric that implements an administrative boundary between many networks prsents unique challenges, from cases such as those outlined above, helping providers learn how to detect and correct cases having default pointed at them, all the way to explaining to Juniper router owners that their "policed discards" counter increments because other participants emit CDP packets (and some of the other non-IP protocols you mention, but most often CDP). More germane to your questioning of Bill's point, though, items such as IPv6 address assignment, and an engineering staff prepared to deal with IPv6-related questions and issues, determine whether an exchange point supports IPv6 above and beyond just letting it happen on their switch fabric. Stephen (now VP of Engineering at PAIX, as well as being a founder, etc.)
In the referenced message, Sean Donelan said:
On Fri, 17 May 2002 bmanning@karoshi.com wrote:
perhaps better late than never... PAIX & LINX both have IPv6 capabilities at/on the exchange fabric(s). I am not aware that Equinix has taken that step.
Uhm, another dumb question.
Why does the operator of a layer 2 exchange care (or know) what protocols your are using? IPv4, IPv6, heck I remember seeing Appletalk, OSI and DECNET on MAE-EAST. What consenting network operators do....
They probably only care on broadcast-based protocols, since is more than the consenting networks. Might also include multicast, which could then get you into the realm of pim snooping, which might then get you up into the ipv4 and ipv6 arena.
On Fri, 17 May 2002 bmanning@karoshi.com wrote:
perhaps better late than never... PAIX & LINX both have IPv6 capabilities at/on the exchange fabric(s). I am not aware that Equinix has taken that step.
Uhm, another dumb question.
Why does the operator of a layer 2 exchange care (or know) what protocols your are using? IPv4, IPv6, heck I remember seeing Appletalk, OSI and DECNET on MAE-EAST. What consenting network operators do....
What step does Equinix (or any other layer 2 exchange) need to do? The ATM NAPs might have an issue due to ATM/ARP, but even then I suspect two consenting network operators could use static IPv6 ARP tables without the NAP operator doing anything.
Two things: IPv6 can and does take advantage of larger MTU sizes. Selection of switch fabric makes a difference. Often, participants expect to have an IP address assigned for their use on an exchange. Usually these delegations are from a common block. Where they are not, its hard to tell an exchange from a bunch of point2point links. LINX and PAIX have IPv6 prefixes that participants can use. I would expect that if the Equinix exchange participants were IPv6 hungry, they would ask for a way to get a v6 address for their connection. And I would expect Equinix would find a way to accomodate them. Otherwise you are correct, the operators don't have to coordinate at all, except on a bilateral basis and then.... whats the point of the exchange? :)
bmanning@karoshi.com wrote:
I would expect that if the Equinix exchange participants were IPv6 hungry ...
Let me toss in a question that may really be dumb... what are those that are hungry for IPV6 doing with it? I figure that organizations that run IPV6 now think they are ahead of the game. Are they? Is this something that responsible ISPs should be doing? Would this turn our network into one big NAT area when we have to translate into IPV4 addresses at the edge to get to the real Internet? -mark !bankrupt, hence !Tier1
bmanning@karoshi.com wrote:
I would expect that if the Equinix exchange participants were IPv6 hungry ...
Let me toss in a question that may really be dumb... what are those that are hungry for IPV6 doing with it?
Id guess meeting customer demand?
I figure that organizations that run IPV6 now think they are ahead of the game. Are they? Is this something that responsible ISPs should be doing?
Depends on what you think the game is. Being able to get more address space than you can conceivably need can be a powerful motivator.
Would this turn our network into one big NAT area when we have to translate into IPV4 addresses at the edge to get to the real Internet?
Perhaps. The migration issues are not all that well thought out and there are some real pitfalls in the details. Call it early adopter syndrome.
-mark !bankrupt, hence !Tier1
Often, participants expect to have an IP address assigned for their use on an exchange. Usually these delegations are from a common block. Where they are not, its hard to tell an exchange from a bunch of point2point links. LINX and PAIX have IPv6 prefixes that participants can use.
or, use link-local address on IX switch as documented in draft-kato-bgp-ipv6-link-local-01.txt. itojun
Often, participants expect to have an IP address assigned for their use on an exchange. Usually these delegations are from a common block. Where they are not, its hard to tell an exchange from a bunch of point2point links. LINX and PAIX have IPv6 prefixes that participants can use.
or, use link-local address on IX switch as documented in draft-kato-bgp-ipv6-link-local-01.txt.
Unfortunatly, that technique does not have broad commercial implementation. In addition, there was some pushback from the IETF IDR wg on using this technique. So its still a bit "iffy", although a very elegent "haq". (we used it in LA for a while but could not get folks to adopt it.)
itojun
--bill
or, use link-local address on IX switch as documented in draft-kato-bgp-ipv6-link-local-01.txt. Unfortunatly, that technique does not have broad commercial implementation. In addition, there was some pushback from the IETF IDR wg on using this technique. So its still a bit "iffy", although a very elegent "haq". (we used it in LA for a while but could not get folks to adopt it.)
cisco IOS can do it, IIRC. i'm always doing it with zebra . anyway, i'm not sure if it fits "broad implementation", so i'll stop. itojun
On Fri, 17 May 2002, Sean Donelan wrote:
On Fri, 17 May 2002 bmanning@karoshi.com wrote:
perhaps better late than never... PAIX & LINX both have IPv6 capabilities at/on the exchange fabric(s). I am not aware that Equinix has taken that step.
Uhm, another dumb question.
Why does the operator of a layer 2 exchange care (or know) what protocols your are using? IPv4, IPv6, heck I remember seeing Appletalk, OSI and DECNET on MAE-EAST. What consenting network operators do....
LINX for example permits very specifically IPv4 only, no multicast including routing protocols etc, no mac broadcasts ie spantree. I think theres a danger on very large switching fabrics that if youre not specific things will happen that are detrimental to all members.. all major switching problems I know of at LINX were caused by members doing something not permitted by the rules... Just because you -could- do something without the operator knowing doesnt mean you should, the rules are there for everyones protection and I think the fact that when people do things they shouldnt it has caused problems speaks for itself in that respect. Steve
What step does Equinix (or any other layer 2 exchange) need to do? The ATM NAPs might have an issue due to ATM/ARP, but even then I suspect two consenting network operators could use static IPv6 ARP tables without the NAP operator doing anything.
On Sat, 18 May 2002 11:14:47 +0100 (BST) "Stephen J. Wilcox" <steve@opaltelecom.co.uk> wrote:
On Fri, 17 May 2002, Sean Donelan wrote:
On Fri, 17 May 2002 bmanning@karoshi.com wrote:
perhaps better late than never... PAIX & LINX both have IPv6 capabilities at/on the exchange fabric(s). I am not aware that Equinix has taken that step.
Uhm, another dumb question.
Why does the operator of a layer 2 exchange care (or know) what protocols your are using? IPv4, IPv6, heck I remember seeing Appletalk, OSI and DECNET on MAE-EAST. What consenting network operators do....
LINX for example permits very specifically IPv4 only, no multicast including routing protocols etc, no mac broadcasts ie spantree.
Doesn't the LINX have a separate LAN for a multicast exchange ? I know that this was set up, but I don't know what it's current status is. Regards Marshall Eubanks
I think theres a danger on very large switching fabrics that if youre not specific things will happen that are detrimental to all members.. all major switching problems I know of at LINX were caused by members doing something not permitted by the rules...
Just because you -could- do something without the operator knowing doesnt mean you should, the rules are there for everyones protection and I think the fact that when people do things they shouldnt it has caused problems speaks for itself in that respect.
Steve
What step does Equinix (or any other layer 2 exchange) need to do? The ATM NAPs might have an issue due to ATM/ARP, but even then I suspect two consenting network operators could use static IPv6 ARP tables without the NAP operator doing anything.
On Sat, 18 May 2002, Marshall Eubanks wrote:
On Sat, 18 May 2002 11:14:47 +0100 (BST) "Stephen J. Wilcox" <steve@opaltelecom.co.uk> wrote:
On Fri, 17 May 2002, Sean Donelan wrote:
On Fri, 17 May 2002 bmanning@karoshi.com wrote:
perhaps better late than never... PAIX & LINX both have IPv6 capabilities at/on the exchange fabric(s). I am not aware that Equinix has taken that step.
Uhm, another dumb question.
Why does the operator of a layer 2 exchange care (or know) what protocols your are using? IPv4, IPv6, heck I remember seeing Appletalk, OSI and DECNET on MAE-EAST. What consenting network operators do....
LINX for example permits very specifically IPv4 only, no multicast including routing protocols etc, no mac broadcasts ie spantree.
Doesn't the LINX have a separate LAN for a multicast exchange ? I know that this was set up, but I don't know what it's current status is.
Yep, its a completely separate LAN operated by LINX.. theres a number of members using it. Actually, I'm not one of them.. I was thinking about this today and wondered if people think they are benefiting at all from using multicast exchange points or even just receiving multicast over say a tunnel. I know the benefits of the technology but in reality, today, is anyone using multicast as an ISP and getting something out of it over unicast? Steve
Regards Marshall Eubanks
I think theres a danger on very large switching fabrics that if youre not specific things will happen that are detrimental to all members.. all major switching problems I know of at LINX were caused by members doing something not permitted by the rules...
Just because you -could- do something without the operator knowing doesnt mean you should, the rules are there for everyones protection and I think the fact that when people do things they shouldnt it has caused problems speaks for itself in that respect.
Steve
What step does Equinix (or any other layer 2 exchange) need to do? The ATM NAPs might have an issue due to ATM/ARP, but even then I suspect two consenting network operators could use static IPv6 ARP tables without the NAP operator doing anything.
On Fri, May 17, 2002 at 11:48:32AM +0200, Iljitsch van Beijnum wrote:
A bunch of us are thinking about multihoming solutions for IPv6. For this purpose, it is useful to know a bit more about how actual networks (rather than the ones existing only as ASCII drawings) interconnect. So:
- What are the 12 - 18 most important interconnect locations in the world? MAE East, the Ameritech, Sprint and PacBell NAPs, PAIX, LINX and AMS-IX come to mind, but from where I'm sitting it's hard to judge whether others are important or marginal.
- To how many of them do typical tier-1 and tier-2 networks connect?
- Using private or public interconnects?
Lets see... Your dead or dieing classics are: MAE East, MAE West, AADS, SprintNap, PBNap, and with luck any other ATM exchanges (and may they turn the power off on any remaining FDDI soon). The biggest exchange point in the US is probably PAIX in Palo Alto, and the biggest public exchange is probably LINX. Both of these are excellent places to be. There are some relatively small regionals like NYIIX where you won't find many large carriers, but they still have their own little nitch markets. For example, you won't find many of the Asian or European networks anywhere else on the east coast, without going to PAIX-PAO or Europe yourself. Then we have "the rest". Equinix and PAIX are both trying to take over the market for exhange points in the US (in similar markets, Equinix = SJC IAD NYC ORD DFW LAX, PAIX += SEA ATL). But without even comparing the quality of services offered, I believe Equinix will be the winner for the following reasons: * Financial stability - Equinix is fairly close to being cash flow positive, while PAIX is fairly close to going to the highest bidder. * Price - In these times of cost conciousness (and transit available for less than the price of peering), many people are taking a step back and realizing that PAIX services are OUTRAGEOUSLY priced vs the competition. Some big carriers are turning down their PAIX switch ports, even at Palo Alto. * Carrier choice - The big carriers have clearly all chosen Equinix as the location for their future peerings (with the exception of AboveNet, who obviously will not set foot in there :P). * More diverse market - While PAIX doubles your rack price if you aren't buying a switch port, Equinix markets to end users looking for carrier diversity (and they have the facilities to support this). This has brought the big carriers to sell transit as well as to peer. Of course each market has their own advantages and disadvantages, for example in Dallas both PAIX and Equinix sit right beside each other at the Infomart. In others, one or the other is out of space. -- Richard A Steenbergen <ras@e-gerbil.net> http://www.e-gerbil.net/ras PGP Key ID: 0x138EA177 (67 29 D7 BC E8 18 3E DA B2 46 B3 D8 14 36 FE B6)
There are some relatively small regionals like NYIIX where you won't find many large carriers, but they still have their own little nitch markets.
There's been rumors of NYIIX and PAIX-NY linking up like SIX and PAIX-seattle.
* Price - In these times of cost conciousness (and transit available for less than the price of peering), many people are taking a step back and realizing that PAIX services are OUTRAGEOUSLY priced vs the competition. Some big carriers are turning down their PAIX switch ports, even at Palo Alto.
Which is why I was surprised that Paul offered PAIX-seattle connectivity for a $300 one-time charge for those who are already connected to SIX. -Ralph
On Fri, 17 May 2002, Ralph Doncaster wrote:
There are some relatively small regionals like NYIIX where you won't find many large carriers, but they still have their own little nitch markets.
There's been rumors of NYIIX and PAIX-NY linking up like SIX and PAIX-seattle.
True, but PAIX-NY is not exactly anything to salivate over.
* Price - In these times of cost conciousness (and transit available for less than the price of peering), many people are taking a step back and realizing that PAIX services are OUTRAGEOUSLY priced vs the competition. Some big carriers are turning down their PAIX switch ports, even at Palo Alto.
Which is why I was surprised that Paul offered PAIX-seattle connectivity for a $300 one-time charge for those who are already connected to SIX.
Good point. Folks running the NAPs have to realize that in this day, you can buy relatively good transit in the $50 to $200/meg range. This makes getting capacity to, colo'ing at, and paying for NAP port cost more than transit, in many cases. IIX is the only exchange point that I've run across that is priced as it should be. -- Alex Rubenstein, AR97, K2AHR, alex@nac.net, latency, Al Reuben -- -- Net Access Corporation, 800-NET-ME-36, http://www.nac.net --
There are some relatively small regionals like NYIIX where you won't find many large carriers, but they still have their own little nitch markets.
There's been rumors of NYIIX and PAIX-NY linking up like SIX and PAIX-seattle.
It's not a rumour. PAIX is interconnecting with NYIIX as soon as the fiber engineering people say that the photons will travel end to end.
* Price - In these times of cost conciousness (and transit available for less than the price of peering), many people are taking a step back and realizing that PAIX services are OUTRAGEOUSLY priced vs the competition. Some big carriers are turning down their PAIX switch ports, even at Palo Alto.
Which is why I was surprised that Paul offered PAIX-seattle connectivity for a $300 one-time charge for those who are already connected to SIX.
We aren't silly, and since it would be silly to fail to recognize that some peers want/need different service levels than others, we recognized it and are acting on it.
There's been rumors of NYIIX and PAIX-NY linking up like SIX and PAIX-seattle.
It's not a rumour. PAIX is interconnecting with NYIIX as soon as the fiber engineering people say that the photons will travel end to end.
Will PAIX be around as an entity capable of providing any services in 3 month?
Which is why I was surprised that Paul offered PAIX-seattle connectivity for a $300 one-time charge for those who are already connected to SIX.
We aren't silly, and since it would be silly to fail to recognize that some peers want/need different service levels than others, we recognized it and are acting on it.
When a company makes such business decisions, would this company be around to make that $300 one time charge worth more than a dinner at Sagami in next 3 month? Alex
It's not a rumour. PAIX is interconnecting with NYIIX as soon as the fiber engineering people say that the photons will travel end to end.
Will PAIX be around as an entity capable of providing any services in 3 month?
PAIX is modestly profitable and has been for years. We are quite healthy. (I cannot answer questions or respond to rumours about the parent company, MFN. PAIX is an "arm's length" subsidiary.)
When a company makes such business decisions, would this company be around to make that $300 one time charge worth more than a dinner at Sagami in next 3 month?
If you're not happy with the service we provide, please let me know or talk to your client services rep.
On 17 May 2002, Paul Vixie wrote:
There's been rumors of NYIIX and PAIX-NY linking up like SIX and PAIX-seattle.
It's not a rumour. PAIX is interconnecting with NYIIX as soon as the fiber engineering people say that the photons will travel end to end.
Listen, I am not trying to be antagonistic, but: Why does this take so long? PAIX-NY is a MFN facility, so, presumably, there is MFN fiber there and ready to go. 25 Broadway is a MFN building, and has been for some time (as in, many people have MFN fiber there). I can't comprehend how this can take so long.
We aren't silly, and since it would be silly to fail to recognize that some peers want/need different service levels than others, we recognized it and are acting on it.
Can you elaborate on this? -- Alex Rubenstein, AR97, K2AHR, alex@nac.net, latency, Al Reuben -- -- Net Access Corporation, 800-NET-ME-36, http://www.nac.net --
It's not a rumour. PAIX is interconnecting with NYIIX as soon as the fiber engineering people say that the photons will travel end to end.
Listen, I am not trying to be antagonistic, but: Why does this take so long? PAIX-NY is a MFN facility, so, presumably, there is MFN fiber there and ready to go. 25 Broadway is a MFN building, and has been for some time (as in, many people have MFN fiber there). I can't comprehend how this can take so long.
I had the same thought. However, it turns out to light a path there's all kinds of climbing down into manholes that has to happen. I'm no fiber expert, but the parent company (MFN) does employ such experts, so let's remain calm. -- Paul Vixie <vixie@eng.paix.net> President, PAIX.Net Inc. (NASD:MFNXE)
In a message written on Fri, May 17, 2002 at 03:24:38PM -0700, Paul Vixie wrote:
I had the same thought. However, it turns out to light a path there's all kinds of climbing down into manholes that has to happen. I'm no fiber expert, but the parent company (MFN) does employ such experts, so let's remain calm.
(Disclaimer, I also work for MFN) I'll echo Paul's comments. While I don't work directly in the fiber side of our business, several things have become clear about NYC: * The streets are crowded. Be that adding conduit, or existing manholes (which are quite full), things are packed in. Anyone who has been in Manhattan can understand this fact. This slows progress. * The city doesn't like you blocking a lane/street. They really don't like it. They have lots of limits on when you can do things. * Virtually all labor is Union. I personally find this neither good nor bad, but it does change slightly how things are done. * Resources are in general in high demand. What I mean by this is that there is enough demand that it's unlikely "extra" resources from a previous activity are still available to support new activity. Sometimes you get lucky, but in general your waiting in line. * Much of NYC is old. This includes the fiber, and conduit. Problems that occur with old things (out of spec, broken when you go back, missing, etc) happen alarmingly often in Manhattan. Also, a general comment on the dark fiber business. Dark fiber means the fiber must be spliced end to end, and until someone comes up with a better idea that means someone has to go out to many manholes to do it. That is the downside of this method of provisioning. I think it's outweighed by the upside, but that's just my opinion. I know nothing about Paul's specific case, but in general be patient with fiber in Manhattan, and from talking to other carriers that is not unique to MFN. -- Leo Bicknell - bicknell@ufp.org - CCIE 3440 PGP keys at http://www.ufp.org/~bicknell/ Read TMBG List - tmbg-list-request@tmbg.org, www.tmbg.org
On Fri, May 17, 2002 at 03:24:38PM -0700, Paul Vixie wrote:
I had the same thought. However, it turns out to light a path there's all kinds of climbing down into manholes that has to happen. I'm no fiber expert, but the parent company (MFN) does employ such experts, so let's remain calm.
Actually, doesn't MFN usually outsource this to Bechtel, Keyspan Communications and others? :)
Actually, doesn't MFN usually outsource this to Bechtel, Keyspan Communications and others? :)
I don't know. It wouldn't change my position on the subject, which is that it's their fiber, and in NYC it's a large and complex plant, and they've got people working on the PAIX/NYIIX path who know a lot more about fiber in general AND this plant in particular than I do. -- Paul Vixie <vixie@eng.paix.net> President, PAIX.Net Inc. (NASD:MFNX)
One thing to remember with New York in general. I have personal experience with dealing with pulling both copper and fiber in NY and its not just technical. Although there are huge problems with pulling new services and fiber on a technical level. Remember that most of the facilities were deployed in the 20's and 30's. Also a large problem is it was my experience pay-offs and some limited bribes were required to have linemen complete their work. A practical example of this was a t1 which I was told would take 3 months to install, after some palms were greaced dropped to 2.5 weeks. New York is a different animal all tgether. On 17 May 2002, Paul Vixie wrote:
Actually, doesn't MFN usually outsource this to Bechtel, Keyspan Communications and others? :)
I don't know. It wouldn't change my position on the subject, which is that it's their fiber, and in NYC it's a large and complex plant, and they've got people working on the PAIX/NYIIX path who know a lot more about fiber in general AND this plant in particular than I do.
participants (26)
-
Alex Rubenstein -
alex＠yuriev.com -
Anthony D Cennami -
bmanning＠karoshi.com -
Brian -
Chris Woodfield -
E.B. Dreger -
Iljitsch van Beijnum -
Jeffrey Meltzer -
Jun-ichiro itojun Hagino -
Leo Bicknell -
Mark Kent -
Marshall Eubanks -
Mike Leber -
Mitch Halmu -
Paul Vixie -
Ralph Doncaster -
ren -
Richard A Steenbergen -
Scott Granados -
Sean Donelan -
Stephen Griffin -
Stephen J. Wilcox -
Stephen Stuart -
Steve Gibbard -
Valdis.Kletnieks＠vt.edu