Talk:Tier 1 network

relevance?

The article goes into detail discussing nomenclature, rationale, and marketing but only briefly mentions technical relevance under 'routing'. However, that mention appears to have a biased view, using terminology such as 'lower tier' when mentioning other networks. The technical irrelevance, or as illustrated in that portion outright danger and detriment to traffic propagation should surface more readily. jzp (talk) 16:05, 20 July 2014 (UTC)

updates

The article title should be "Tier 1 ISPs". A tier one carrier, meaning a telecom carrier of voice services is something different and is definable via FCC rules. —Preceding unsigned comment added by 208.46.68.194 (talk) 2005-10-21 14:01:08 (UTC)G

"This is what happened between Cogent and Level 3 recently." --This needs to be replaced with the date of the occurance and a brief explanation. I know Wikipedia tries to be timely but "recently" ages fast. I am leaving this comment as I am not the one qualified to write about this. Ray Trygstad 14:27, 21 October 2005 (UTC)

this appears to have been fixed. Good call. Jasongetsdown 17:23, 21 October 2005 (UTC)

Should be locked and/or have an accuracy dispute disclaimer: In light of the persistent vandalism (some of it traceable to Cogent, and I don't use the term vandalism lightly), shouldn't this article be appropriately marked with a disclaimer that its accuracy is in dispute and locked so that non-registered users can't edit? Same comment has already been made in talk for Cogent Communications —Ke4djt @ 1358, 04 May 2006 (UTC)

http://en.wikipedia.org/wiki/Talk:VSNL_International_Canada VSNL's article seems to describe it as a Tier1 ISP. Either that article or this needs fixing. —Preceding unsigned comment added by 24.141.255.13 (talk) 2006-06-16 03:54:21 (UTC)

"Teleglobe/VSNL AS6453 does not buy transit from 1239. Easily verifiable in the route maps. Changed the label accordingly" Triviacontributor 05:10, 13 December 2006 (UTC)

Proposed merge from List of tier 1 internet service providers

The companion article is just a short list, and really has no meaning outside the context of the main article, and isn't referenced outside the main article. -- Randal L. Schwartz 16:56, 30 October 2005 (UTC)

just because you put an article on wikipedia doesn't mean the information is true. If you ask 100 people what a Tier 1 ISP is you'll get 10+ different answers...there is no common definition. Just because you don't own the fiber in the ground doesn't mean you don't privately peer with the other larger ISPs. —Preceding unsigned comment added by 24.147.208.157 (talk) 2005-11-21 15:32:45 (UTC)

Merging List of tier 1 internet service providers

I'm merging the contents of List of tier 1 internet service providers. This should be a quick merge.

However, that article is in the odd situation of having a talk page MUCH longer than the article itself! I don't know if we're supposed to merge the talk pages too. For now, here's a link to the talk page as it stands now... talk page snapshot Jamie 01:33, 14 November 2005 (UTC)

Merge complete. The page still needs to be cleaned up and wikified. Jamie 01:41, 14 November 2005 (UTC)

hah?

It reads "Tier 1's gain a significant portion of their IP Transit revenue from traffic which stays "on-net", by being delivered between two customers without ever leaving its network." Since they work only on peering isn't "free" delivery the _only_ thing tier 1 carriers know? --161.76.99.106 00:55, 23 March 2006 (UTC)

The reason this is the case is becasue Tier 1 providers actually get paid _twice_ for on-net traffic as it travels between customers instead of going from a customer to a peer. --- Jwvo 04:19, 19 August 2006 (UTC)

It's still an insignificant amount. The largest it's _ever_ been for _any_ provider, at any time in history, was 0.65%, for UUNet, just before their bankruptcy. I don't think 0.65% constitutes a "significant portion." Bill Woodcock 03:40, 31 January 2007 (UTC)

Uhh, Bill...what? Sprint, just a few years ago (2002/2003) was doing about 30% "on-net". It's definitely decreased, but a number of the Tier 1's still have signifigant "on-net" traffic.

—Preceding unsigned comment added by 70.198.99.224 (talk) 2007-03-31 21:00:43 (UTC)

List of tier 1's updated

List of tier 1 isp's were updated, to include several that people always leave out as well as fix the status of Cogent Communications. Which is _NOT_ a tier 1 isp. Soms 2006-05-03 15:10:19 (UTC)

Since folks wanted to explicitly exclude one ASn from one of thee entrants, fleshed out several of the other ASNs from the named entities which 'don't count". jzp 2006-05-03 17:52:43 (UTC)

Someone added Teleglobe/VSNL to the list of Tier 1s. I fixed the entry to match the others and updated the total number at the top of the list. I did this only to fix the readability of the list and I can't vouch for the accuracy of including Teleglobe/VSNL on the list. If someone with the expertise can do a check I would greatly appreciate it. Sperril 19:01, 1 November 2006 (UTC)

Nevermind, it was removed. Sperril 19:40, 2 November 2006 (UTC)

Added XO Communications to the List of Tier 1 ISP's. XO Opperates a 9 Core Node OC-192 Backbone, is publicly peared at all 7 traditional NAPs and at the two new NAPs that have been created in Miami (NAP of the Americas) and the Seattle NAP. XO leases no transit service and has a 100% default free back bone with over 220 private peering points in 9 domostic and 3 international locations. —Preceding unsigned comment added by 205.158.160.209 (talk) 2007-02-23 20:35:03 (UTC)

seabon is become telecom italia sparkle ad it is tier 1 — Preceding unsigned comment added by 80.117.72.174 (talk) 17:26, 14 August 2013 (UTC)

NPOV Tag

The tag was added by MureninC, who (from text not shown correctly in the tag) says:

"This list looks really strange, it only has one non-american company, and major players like Telia are not here at all; information must be extended, and if Telia etc. do not qualify, then they must be put into Tier 2 article with explanations"

I don't know enough about the subject to comment either way, so comments are appreciated. It would probably help to have verified sources for this information (cf. WP:NOR) --H2g2bob 16:56, 13 June 2006 (UTC)

Changed to globalise tag, as that seems more in-keeping with the problem --H2g2bob 17:12, 13 June 2006 (UTC)

Moved text by 81.234.199.232 from page to here (info appears to be correct, so adding): --H2g2bob 23:25, 18 June 2006 (UTC)

Question, is TeliaSonera AS1299 a tier 1 carrier?

YES:

• Renesys names TeliaSonera AS1299 a tier 1 network/carrier. Renesys is a solid player in this field.
• The previous gw that was mentioned does no loinger exist/being used, the customer part is also removed. From what we can see TeliaSonera AS1299 is now using two major interconnection points, one in San Jose and the other on in New York. Not showing evidence of transit.

NO:

• According to Verizion AS701 peering paper, TeliaSonera does not qualify for peering. Specifically, Verizon requires that a settlement-free peer meet them in all major cities where they share a mutual presence. Telia and Verizon only appear to interconnect in New York, Ashburn, and San Jose; they do not connect in Chicago, Dallas, Miami, Los Angeles, et cetera.
• Verizon maintains a strict architecture in which peers (settlement-based or settlement-free) are connected on border ("BR" in traceroute) routers, and paying customers on gateways ("GW"). Telia only connects to "GW" routers.

Source for this?! —Preceding unsigned comment added by 83.227.136.194 (talk) 2008-06-11 20:49:53 (UTC)

• According to Level(3) BGP Communities describtion: [1]

  Communities allow suppress or prepend to peer AS, where
     - peer AS has a peering connection to Level 3
     - peer AS is not a customer of Level 3
  65000:XXX - do not announce at peerings to AS XXX

prefixes with 65000:1299 community set should suppress at peerings to AS1299 (TSIC), but ignore if TSIC is customer of Level(3). During my research at January 06, 2009 Level(3) ignores 65000:1299 tags and prefixes are appear in TSIC network as is. —Preceding unsigned comment added by 77.120.63.58 (talk) 21:33, 6 January 2009 (UTC)

additional discussion follows, though this should be cleaned up into pros and cons listed above by somebody...

What about TeliaSoneraIC AS1299? (not AS 5518, 3308 or 3301) --81.234.199.232

For the record, 70.224.205.218 removed this again, as "Telia is not a Tier1 (see def: settlement free and no transit)" --H2g2bob 16:43, 21 June 2006 (UTC)

Would someone like to tell me why TIER1-carrier Telia IC 1299 is being removed? —Preceding unsigned comment added by 81.234.199.232 (talk • contribs)

Re-added TeliaSoneraIC with disputed tag. The crux of this confusion, I think, may stem from the definition of Tier 1 carrier. TeliaSoneraIC claims on it's website that it's a "Tier 1 carrier" [2], but it may be unclear what definition of "Tier 1" they are using. Tier 1 on this page is stated as being an ISP which doesn't pay for peering with any other ISPs. I don't know any way of verifying whether the ISP is Tier 1 in this context or not, perhaps someone can suggest something. --H2g2bob 19:09, 22 June 2006 (UTC)

PS, Don't forget you can sign posts like this: ~~~~

I havent got any ideas of TSIC peering agreements, but they have directpeering to all other TIER1-carriers in market, and they are a very significant provider in Europe. When AGIS was bought several years ago, they "received" TIER-1 status in the US(which I guess they already had in Europe). ~~81.234.199.232

why has TSIC's been removed again(excuse my language now) what the f*ck? If someone's gonna keep removing it, It would be pretty good if that person had some arguments stating that TSIC is not TIER1, if not stop removing it. —Preceding unsigned comment added by 81.234.199.232 (talk • contribs)

Reverted edit so it's back on the page. It has a disputed tag, so it's bad wikiquette to remove it without explanation on the talk page. --H2g2bob 11:32, 24 June 2006 (UTC)

TSIC removed again.... I'm still trying to find an explanation. Couldnt anyone lock the edit-function so its not possible to just remove without a valid explanation? —Preceding unsigned comment added by 81.234.199.232 (talk • contribs)

Why? Simple. Telia purchases transit from 701 and 7018 (and others?) and is therefore not a tier 1 provider. —Preceding unsigned comment added by 68.193.227.33 (talk • contribs)

Still just plain comments without any resources given at all. And I can tell you one thing for sure - that is tsic does NOT purchase transit from 7018. —Preceding unsigned comment added by 81.234.199.232 (talk • contribs)

No, TSIC purchases transit from AT&T and UUNET. This is easily verifiable w/ BGP communities and looking at PTR records for the kind of devices they connect to. Would the fact that you're posting from a telia.com host (corporate firewall?) have anything to do with your inaccuracy and bias? :) —Preceding unsigned comment added by 68.193.227.33 (talk) 2006-06-29 20:55:39 (UTC)

I'd like to hear how you can determine who x buys transit from by looking at PTR records? —Preceding unsigned comment added by 81.234.199.232 (talk) 2006-06-30 17:04:09 (UTC)

teliasonera.CUSTOMER.alter.net == probably a customer and not a peer, one would think :) —Preceding unsigned comment added by 68.193.227.33 (talk) 2006-07-02 05:55:01 (UTC)

Sure, I've seen that one, but I havent seen one from AT&T, unless you are thinking of att-gw.ny.telia.net, though that one doesnt reveal any peer/transit. —Preceding unsigned comment added by 81.234.199.232 (talk) 2006-07-02 11:36:00 (UTC)

You know guys, we could leave this discussion at "they purchase UUNET transit and are therefore not settlemen free" and end the bickering right now. TSIC is not tier 1. —Preceding unsigned comment added by 68.39.139.18 (talk) 2006-07-03 15:25:57 (UTC)

---

Wrong! TSIC is an tier1 provider. It has the biggest and strongest european backbone, it does not buy traffic from other providers in north america, simply does a traffic exchange. Therefore its a tier1 provider and should be listed as one.

Edit2, Telia is not TeliaSonera International Carrier, Telia is the local provider for Sweden. Get the fact straights. —Preceding unsigned comment added by 83.227.136.194 (talk) 13:20, 4 October 2007 (UTC)

---

Also I wish to add, that Russian branch TSIC purchases the Russian transit from company TTK (as20485). Here so for example from the Moscow part of network TSIC it is visible company GLDN.NET, which is peer of company TTK.

Edit: No you donkey, they buy transit from tsic, you notice the c before telia.net, that c stands for customer. Pay for russian transit, what a joke. —Preceding unsigned comment added by 83.227.136.194 (talk) 11:58, 12 April 2008 (UTC)

1 transtelecom-113434-mow-b2.c.telia.net (213.248.97.174)

2 TTK-lgw.Moscow.gldn.net (194.186.0.193) [AS 3216]

3 cat03.Moscow.gldn.net (195.239.10.189) [AS 3216]

195.209.41.226 22:01, 25 August 2007 (UTC)Russian user

List of Not-Tier-1's

Ok people, I know we ALL think that our favorite carriers are Tier 1's because their website says it is, but that doesn't make it so. In addition to some blatant vandalism from Cogent and Telia staff repeatedly adding themselves, there are no shortage of well meaning users who want to add networks because they legitimately believe that they know of a new tier 1 which hasn't been mentioned. Please, let me assure you that every true Tier 1 has been included in the list of 9, and all of the common misconceptions and previous subjects of debate have been added to the Not-Tier-1 list, along with a verifiable reason why they are not.

Just because a network is not a Tier 1 does not make it any more or less important, so PLEASE stop vandalizing the legitimate technical facts of this article with marketing. If you think you have a network which is a tier 1 and isn't listed, please ask here and someone will be happy to research it and tell you why it isn't and help you add it to the Not-Tier-1 list. If you have any doubt in your mind, ask yourself "am I an expert in the complex issues surrounding tier 1 interconnection?". If the answer is no, the network you're thinking of is probably not a tier 1.

Thanks you! :)

Humble226 07:58, 8 July 2006 (UTC)

I guess youre referring to me(81.234.199.232) as a "telia employee", I would just ask you to do a lockup or a check on ripe, which will verify where and to what that ip-adress belongs to. HINT - it's not 1299. You mentioned research, obviously you missed that part yourself. —Preceding unsigned comment added by 81.234.199.232 (talk) 2006-08-09 21:56:40 (UTC)

AboveNet does NOT buy from Sprint where do you people get your information from. Take your sales FUD and go sell POTS lines. —Preceding unsigned comment added by 64.124.244.12 (talk) 2006-10-05 14:52:02 (UTC)

AboveNet does not pay for peering with Sprint they pay a fines for bad peering ratios. Does this still put them as a non-tier 1 provider? —Preceding unsigned comment added by 208.184.57.9 (talk • contribs) 00:28, 26 April 2007

We need to take a look at AS6461 (abovenet)... I can't find any evidance that they actually buy transit as stated here. In fact, until 2004 they provided transit for cogent. If i take a look at a BGP feed from them, I see every tier one as a direct neighbor. That being said, if they are tier 1, they are the smallest one given they only announce about 4700 routes to their peers --- Jwvo 03:28, 19 August 2006 (UTC)

Under normal circumstances they use BGP communities to only accept Sprint customer routes, and to restrict propagation of their announcements to Sprint customers. But they definitely receive a full table from Sprint, and they have used it in the past when necessary. For example, when ATDN was threatening to depeer them due to their ratio, the Sprint routes came out to reduce traffic via ATDN peers. You can also see that they are terminated onto customer gateway routers, not peering routers:

Name: sl-gw19-rly-3-0.sprintlink.net Address: 144.232.247.85

Name: sl-abovenet-19-0.sprintlink.net Address: 144.232.247.86

Until recently you could even see the full routes in their looking glass, but they finally blocked access to this view.

Humble226 18:02, 22 August 2006 (UTC)

It looks like cogent just recently started getting a direct peer from sprint. The router names on the sprint traceroute are identical to those on Sprint's Level3 peer. Based on that, I am going to assume it is settlement free:

trace from sprint to level3:

 1 144.232.1.170 0 msec 0 msec 0 msec
  2 sl-st20-la-11-0.sprintlink.net (144.232.8.94) 108 msec 4 msec 196 msec
  3 so-1-0-0.gar1.LosAngeles1.Level3.net (64.152.193.73) [AS 3356] 4 msec 108 msec 12 msec
  4 ae-2-52.bbr2.LosAngeles1.Level3.net (4.68.102.33) [AS 3356] 0 msec

trace from sprint to cogent:

  1 sl-bb24-ana-13-0.sprintlink.net (144.232.1.138) 0 msec 0 msec 0 msec
  2 sl-st21-la-13-0.sprintlink.net (144.232.20.69) 0 msec 4 msec 4 msec
  3 154.54.13.41 [AS 174] 72 msec 0 msec 0 msec
  4 t3-2.mpd01.lax01.atlas.cogentco.com (154.54.3.21) [AS 174] 4 msec 4 msec 0 msec
  5 t8-4.mpd01.iah01.atlas.cogentco.com (154.54.2.197) [AS 174] 36 msec 32 msec 36 msec

Cogent does however seem to be continuing to use verio transit to reach AOL (traceroute from cogent's side):

Type escape sequence to abort. Tracing the route to www.aol.com.websys.akadns.net (64.12.192.2)

 1 g10-0-224.core01.lax01.atlas.cogentco.com (66.250.4.5) 4 msec 4 msec 0 msec
 2 t3-1.mpd01.lax01.atlas.cogentco.com (154.54.2.102) 4 msec 4 msec 0 msec
 3 t2-1.mpd01.sjc01.atlas.cogentco.com (154.54.5.181) 16 msec 12 msec 16 msec
 4 v3490.mpd01.sjc03.atlas.cogentco.com (154.54.6.82) 12 msec 16 msec 12 msec
 5 verio.sjc03.atlas.cogentco.com (154.54.10.50) 36 msec 12 msec 12 msec
 6 p16-0.aol.snjsca04.us.bb.gin.ntt.net (129.250.9.42) 16 msec 12 msec 16 msec
 7 bb1-sjg-P0-0.atdn.net (66.185.150.80) 12 msec 16 msec 12 msec
 8 bb1-ash-P14-0.atdn.net (66.185.153.58) 84 msec 84 msec 84 msec
 9 pop3-ash-P0-0.atdn.net (66.185.148.209) 188 msec 84 msec 84 msec
10 dar1-mtc-S0-0-0.atdn.net (66.185.148.222) 84 msec 84 msec 84 msec

Jwvo 18:06, 31 October 2006 (UTC)

List of Not-tier-1's

I do think the 'not tier 1' section is quite confusing. I understand the need for a list of those commonly (and wrongly) assigned to being Tier 1 but my first impression when scrolling through the article was that this must be a list of Tier1's, now the text does make this clear but I think maybe a heading with something like "Non Tier 1 networks attributed to be Tier 1" (or something slightly more concise perhaps?)

ny156uk 08:04, 5 August 2006 (UTC)

Is ATDN a Tier 1 Network?

I am looking at Route views routing archive to test tier 1 claims. I am a very beginner. When I saw the following, my initial thought was that ATDN (AS1668) buys transit from 703. Isn't that true? If not, can someone tell me which ISPs are peering and which ISPs are buying transit from others?

 [vijay@lab data]$ grep "2914 701 703 1668" oix-full-snapshot-2006-07-01-0000.dat | uniq
 *                   129.250.0.85             6             0 2914 701 703 1668 i
 [vijay@lab data]$ grep "7018 701 703 1668" oix-full-snapshot-2006-07-01-0000.dat | uniq
 *                   12.0.1.63                              0 7018 701 703 1668 i

--Vijaykcm 04:28, 25 August 2006 (UTC)

Who was forgotten?

Whoever did this list seems to have forgotten Cable & Wireless. They are a British company who owns, operates and leases backbone, pipe, bandwidth (whatever you want to name it) to other telecom companies across the world. To leave them out of a list like this seems to be a disservice by the uninformed. —Preceding unsigned comment added by 170.35.224.64 (talk) 2006-08-24 22:52:47 (UTC)

---

Cable & Wireless (AS1273) buys transit from Level 3 (AS3356) and Savvis (AS3561). You can see paths like the following ones from route-views.

 11608 2914 3356 1273 8866 9070 20876 31296 35424
 11608 2914 3356 1273 9158
 11608 2914 3356 1273 9158 15564
 11608 2914 3356 1273 9158 35376
 11608 2914 3561 1273
 11608 2914 3561 1273 10292
 11608 2914 3561 1273 1103 1104
 11608 2914 3561 1273 1103 1128
 11608 2914 3561 1273 1103 12654

--Vijaykcm 04:46, 25 August 2006 (UTC)

XO a Tier 1

I noticed that XO was recently added as to the list of true tier 1 networks. Can anyone confirm that they are no longer subject to being paid peers of Sprint and Level3, as the previous version of this page indicated? —The preceding unsigned comment was added by Jaufder (talk • contribs) 20:38, 12 March 2007 (UTC).

...Added, by someone from behind an XO netblock. Draw your own conclusions. —Preceding unsigned comment added by 82.38.130.54 (talk) 2007-03-13 22:02:56 (UTC)

Telstra?

What about Telstra, AS1221? Who is their upstream provider? It seems most routes go via AS4637 which is Reach, half owned by Telstra. I'm not an expert in this matter, I'm looking for information. Icd 02:49, 23 March 2007 (UTC)

Telstra have definitely extended their reach of late it seems, interconnecting (albeit cannot confirm if its paid peering or settlement free peering) with at least 1/2 of the existing Tier 1 Networks... this seems to be the strategy taken since Reach was renamed Telstra Global after Telstra bought out PCCW's 50% stake... AS4637 now seems to be going after that fully interconnected status, looking at their global topology nowadays. Whether they will meet true Tier 1 Network status is yet to be confirmed... once the last remaining networks are peered etc. — Preceding unsigned comment added by 124.170.133.11 (talk) 10:08, 28 December 2012 (UTC)

Bad

The wiki article is poor. There is no "backbone" and there are no pure Tiers. Most networks are various mixes of paid and settlement-free peering at private and public exchanges. Useful distinctions might be made among networks by their composition of the mixtures. A useful network quality comparison might include the number of routes and latency to various networks, but such latency measurements would be subject to biases dependent on where the sensors were placed. (Whichever networks the sensors are located in will measure better than others.) If we assume most of the large networks are internally similarly fast due to being built with big pipes roughly comparable to other large networks, then the number of (aggregated) routes they can provide may be a useful comparison criteria. Given reasonably symmetrical peering agreements, more routes generally means more direct connections. Having more direct connections between similarly fast networks will generally result in the lowest latency with the other networks. —The preceding unsigned comment was added by 209.204.169.179 (talk) 13:30, 13 April 2007 (UTC).

Regardless of the usefulness/uselessness of the tier definitions (and you're probably right), the article describes how the term "Tier 1 ISP" is used and understood, it is not the original source of the definitions, so your description of the article as "bad" isn't quite appropriate. Thedangerouskitchen 15:37, 17 April 2007 (UTC)

Fair enough, but meta-discussions about the appropriateness or background assumptions of a given article probably constitute meaningful commentary. —The preceding unsigned comment was added by 209.204.169.179 (talk) 10:38, 21 April 2007 (UTC).

This article is of use to those trying to understand what it means to be connected to various networks with respect to hosting decisions. If I choose to buy hosting/networking from a tier 1 company I know that they are not going to loose service to parts of the internet because they didn't pay their bill. They *might* loose connectivity because of a peering dispute but that's an entirely different level of risk. If I buy hosting/networking from a tier 2 provider there is the chance that I would loose access to parts of the internet due to them not paying their bill or due to a peering dispute, higher level of risk and different ways the risk could manifest itself. If I buy service from a tier 3 provider I have the risk of loosing access to all of the internet because they didn't pay their bills. Additionally peering agreements are different from transit agreements in ways that could affect end customers differently (different circumstances in which one network will accept/deliver traffic to another). - mec 12 August 2007

I tend to think that a tier 2 network is more reliable in this regard than a borderline tier 1. A network that thinks of itself as tier 2 is going to be in the habit of paying bills, and good ones typically buy transit from at least two tier 1 providers. If there's some temporary billing screwup, they'll probably still have full transit from their other tier 1 during the brief loss of one upstream while the billing problem is resolved. Whereas a carrier that thinks of itself as tier 1 that isn't thought of as a tier 1 by the bigger network operators may be more likely to get into a fight where it refuses to pay money, and its customers lose out. JNW2 19:47, 2 September 2007 (UTC)

"If I choose to buy hosting/networking from a tier 1 company I know that they are not going to loose service to parts of the internet because they didn't pay their bill." It's true that a tier 1 won't lose connectivity due to not paying a transit bill but there are plenty of other bills they need to pay to keep the service going so I don't think this is very relatvent.

The trouble with using a tier1 as your sole source of connectivity is that your your connectivity to single homed customers of other tier 1 providers is reliant on a single peering relationship between the tier 1 you are using and the tier 1 the person you want to communicate with is using. If that relationship goes sour then you will lose connectivity to some chunk of customers. OTOH as JNW2 says a moderate to large tier 2 will likely buy transit from multiple tier1 providers so losing one peering or transit relationship should not cause protracted loss of connectivity (it may cause temporary loss while routes update). 130.88.108.187 (talk) 13:14, 18 June 2012 (UTC)

Peering Agreement Standards

Are there standards which a "Peering Agreement" must meet in order to be considered a peering agreement for the purposes of this article? Are there different types of Peering Agreements? For example if three networks all interconnect (A, B and C) could network A have an agreement with B that says they will relay to C in the event of a failure in the A-C connection but at the same time have an agreement with C that says that they will only peer for traffic between A and C but will not pass on traffic to B. If so should these be examined and tiered? mec 12 August 2007

If A is a tier 2 network, C is a tier 2 network, and B is a tier 1 network, there are probably plenty of examples of that in the real world. In the typical case, A is buying transit from B, and C is buying transit from B, and there's some more direct and cheaper A-C settlement-free interconnection. JNW2 19:55, 2 September 2007 (UTC)

Clean Up Badly Needed

This article is currently incoherent. I understand that some of the content in earlier versions was controversial, but the current version of the article is just a mess. —Preceding unsigned comment added by Geoffsnowman (talk • contribs) 23:27, 2 October 2007 (UTC)

I don't understand what has been written here, and I am not any closer to understanding what a 'Tier 1' network is. If was a able to, I would delete this entire article and start from new. Schnarr 07:54, 3 April 2009 (UTC)

Is it possible to prove which networks are a Tier One?

It has been noted that the "facts" in this page are unsupported. This is why I wrote:

  Unfortunately, it is impossible for an outside authority to confirm
  that a network is not paying settlements of any type because such
  business agreements are frequently not public information, or even
  covered under a Non-Disclosure Agreement.

The peering community is small, a few hundred people - less for networks which span continents or oceans. The community knows some things, and is unsure about others. The seven tier one networks listed really, really are tier one networks, and you will not find an authoritative source for that information other than people who know. The two which are in question are actually believed to still be tier one, but because it is not "known" for certain, I listed them as possibles.

Transit free is a totally different story. That can be objectively confirmed from the outside, and is trivially done. So there is no question about the 12 networks under discussion. The only question is what business arrangements they have, and that is supposedly secret. But like any good secret, it leaks.

What we are trying to present here is some information for the rest of the world to see. If you don't believe it, no problem, it really doesn't matter anyway.

-- ianai July 2, 2008 —Preceding unsigned comment added by Ianainet (talk • contribs) 05:19, 2 July 2008 (UTC)

A few points: First, it's a poor use of the discussion page to place a new comment at the top with no header or anything to make it clear what is being discussed. Second, it is completely inappropriate to both the spirit and the guidfelines of wikipedia to represent the opinions of a posited group (the "peering community") as fact. You admit that the edits you have made to this article are based on uncited and unverifiable sources, and yet you wish these contributions to be treated as fact. That's just not how wikipedia (or any encyclopedia, for that matter) works. If transit-free is verifiable, fine, verify that, but attempting to then split hairs over which of those companies may or may not have paid agreements with each other is a fool's game. -Dayv (talk) 21:36, 2 July 2008 (UTC)

Thinking further on this, there are two verifiable pieces of information we can base any classification on: which networks are transit-free (verifiable via public route-servers), and which networks are marketed as "Tier 1" (verifiable through citation of marketing documents). As the Tier 1 designation is entirely a marketing construct anyway, you will really never be able to do any better than this. There may be some alternate world where the definition and verification of Tier 1 status is made by some arbitrary controlling body, but we do not live in that world and are better off documenting the world we do live in. -Dayv (talk) 21:55, 2 July 2008 (UTC)

Sorry if I have broken protocol. This is my very first edit of a discussion page, and I didn't see docs on how to do it properly. Reading the rest of the page, it didn't look like there was much in the way of protocol. I have added the appropriate header. Please let me know if I missed anything else.

As to whether the facts are supportable or not, this is not a peer reviewed journal nor an encyclopedia (despite the suffix). The information is clearly presented as the near unanimous belief of the group of people who are the only "experts" in the field, not some objectively verifiable fact. Put another way, this is objective fact: The fact that the overwhelming majority of peering coordinators for large networks believe this is not in question. Whether that belief is objective "truth" can be debated, but the fact of the belief cannot. Are you arguing this fact be withheld from the Internet?

Lastly, I would strenuously object to listing networks as Tier One because their marketing materials claim they are. There are non-transit-free networks who claims to be "tier one" in their marketing material. Also, Tier One is not simply a marketing construct. The definition presented is rigorous and historically accurate. The fact there are political / legal issues which hide some of the necessary evidence does not mean the definition is insufficient. If you have an alternate definition, or are uncomfortable with the definition presented, you are welcome and encouraged to write those up and add them to the page. That doesn't make the definition presented any less valid, just the opposite.--ianainet (talk) 15:35, 5 July 2008 (UTC)

Provability is, in fact, at the heart of wikipedia's mission. Facts listed in wikipedia should be verifiable, should not be original research, and should be presented with a neutral point of view. These are Wikipedia's three core content policies. As far as verifiability goes, the policy I linked to clearly states that "the threshold for inclusion in Wikipedia is verifiability, not truth—that is, whether readers are able to check that material added to Wikipedia has already been published by a reliable source, not whether we think it is true."

Even outside of the bounds of these policies, the information in this article is not correct. I know for a fact that one of the networks listed here as a tier 1 network pays for peering with one of its transit-free "partners." However, I am not adding this information to the article because I can not cite a source for others to verify this. To do so would be counter to Wikipedia's most basic policies. Wikipedia is not built on the basis of "trust me, it's true," no matter how much some editors might wish that it were.

As to whether or not Wikipedia is an encyclopedia, that is in fact its purpose, both stated and actual. Check Wikipedia if you don't believe me.

Returning to the discussion of this article I posit the following: the only appropriate solution is to list those providers which describe themselves as tier 1, are transit-free, and cannot verifiably be shown to pay for peering. -Dayv (talk) 22:26, 7 July 2008 (UTC)

No Free transit agreements?

The article seems to assume that free transit agreements could not exist (I doubt they exist in reality for business reasons, but the business could potentially change such that these would make economic sense.) [For the following, assume that all mentioned peering is charge-free.

This would be an agreement in which A and B are peered, as are B and C, and as part of this arrangement B will provide free a->c transit and c->a transit. Thus A and C are not directly peered, but packets on Network A could reach C without charge. Assuming A, B, and C are all peered to any other Tier-1 providers, then A and C would reasonable need to be considered Tier-1, as packets from either could reach each other and all other Tier 1 hosts (and thus the entire internet) without incurring any charge.

Thus the statement "Therefore, in order to be a Tier 1, a network must peer with every other Tier 1 network." which is found in the article is not true.

As to how this could happen, imagine if B was a Tier 2 provider who desired to become a tier-1 provider. They managed to get peering arangements with all Tier-1's except A and C (who where peered at the time with each other). These two were refusing to reach a peering agreement with B. However, B decides to offer free transit to each other through its own network. B offers this because B feels it might help reach peering agreements with A and C. A and C agree because their relationship is currently strained, and this takes the pressure off, as it won't prevent connections if they de-peer. Obviously A and C may decide to de-peer, being sick of dealing with each other. B has become an intermediary. Thus as far as A cares, all of C's traffic is actually just B's traffic, with the reverse also being true.

Unlikely? Sure! but plausible enough to destroy the concept that Tier 1's must peer with all other Tier-1's. 24.185.237.31 (talk) 05:01, 6 July 2008 (UTC)

Interesting point, and one that, to be honest, I had not considered. However, There are many things which are strictly not "impossible" which will never happen.

For instance, in your example, by accepting B's offer, networks A and C lose their Tier One status and their Transit Free status - both of which are jealously guarded. Additionally, they promote B to both Transit Free and Tier One. No network is going to destroy their market power and elevate a rival. So while I guess it is not impossible, I am believe this is so far out of the realm of probability that we can leave it as an interesting, but nearly impossible, edge case.--ianainet (talk) 21:08, 6 July 2008 (UTC)

Savvis

Is there really controversy around Savvis' Tier 1 status? There is no citation of source on that claim, and they are commonly regarded as a Tier 1 amongst most I speak to in the ISP community. Toph (talk) 20:50, 24 September 2008 (UTC)

This article is full of unverifiable and controversial content as well as a great deal of what could charitably be called original research. A full rewrite is needed, but I have largely bowed out of this attempt because I work in the industry myself (I feel I can be neutral, but others can easily disagree) and I felt I was edging up to the possibility of an edit war with another user. You can find some of our debate on this elsewhere on this discussion page. -Dayv (talk) 20:17, 30 September 2008 (UTC)

data on who buys from who

Is there any way to find out who an ISPs upstream transit providers are other than asking them? —Preceding unsigned comment added by Plugwash (talk • contribs) 2008-11-03 22:22:46 (UTC)

Define transit free

Could someone please define "transit free" ? It's surprising that it's not defined in the article. Leotohill (talk) 00:41, 1 March 2009 (UTC)

Less space on definitions, more on what they do

I just stumbled on this article from UUNet. After reading it I don't think that I have any better sense of what a tier one provider is. So much of this entry is just trying to define a tier one provider, and going back and forth about who could be considered one. The title of this entry isn't "List of tier one providers," so it should be explaining what they are and what they do. I think the opening sentence should be expanded to give a better idea of what happens at the tier 1 level. The long definition section isn't particularly valuable, in the opinion of this single reader. —Preceding unsigned comment added by 24.7.105.107 (talk) 15:54, 29 April 2009 (UTC)

as to AS

Since AS is an abbreviation of autonomous system, should it not be spelled 'AS' and not 'as', as is currently written all over the article? Reading http://en.wikipedia.org/wiki/Abbreviation, it is not entirely clear, but I believe that AS would be the preferred use. I will go ahead and change if no one objects!? Kll (talk) 05:36, 19 August 2009 (UTC)

My feeling is that 'AS' is the correct abbreviation. 'as' is just improper capitalization, I think. This AS request form from ARIN uses "AS Number" in the text. Go ahead and make your proposed change without fear! Cheers, —fudoreaper (talk) 17:52, 19 August 2009 (UTC)

A tad late, but I've now gone ahead with this change :) Kll (talk) 18:33, 20 November 2010 (UTC)

as6762

The claim (transit free) is verifiable by those of you who have an account on Renesys' MI tool. Pf 193.0.24.206 (talk) 09:18, 16 November 2010 (UTC)

Level3/Comcast

As seen in their press release (http://www.level3.com/index.cfm?pageID=491&PR=962), Level 3 is at least temporarily paying Comcast to serve content to Comcast users. Some users (or user) have taken this to mean that they have a paid internet transit agreement with Comcast, qualifying them as a Tier 2 provider instead of a Tier 1. I believe this is false for several reasons.

First of all if you look at the definition of Tier 1 it is as follows, "tier 1 network is one that can reach every other network on the Internet without purchasing IP transit or paying settlements. By this definition, a tier 1 network is a transit-free network that peers with every other tier-1 network."

Comcast is a Tier 3 network. Level 3 can reach Comcast's network without ever paying a fee (as per the Tier 1 definition). The issue is that Comcast now wants to charge to forward data on its network to its USERS. A Tier 1 network must reach all other networks for free, not every user. For example if I create a private network (and connect it to the internet) and then require anyone to pay to access the computers on the network, this doesn't cause all Tier 1 networks to suddenly become Tier 2 because they can't reach my networks users unless they pay a fee. The Tier 1 networks can still reach my network for free, it's just a closed network and if they want to reach the individual computers there would be a fee.

Second, this isn't a transit agreement. Transit refers to my network A, using network B to reach network C. Network A in this case would be Level 3, network B is Comcast, network C doesn't exist. The traffic that Comcast wants to be paid for is traffic that ends on their network. This isn't how the internet works and is unprecedented for a Tier 3 network to charge to have data from the internet delivered to it while also charging your users to have it delivered. If anything, you pay to have data delivered to your Tier 3 network.

Third, if you want to get technical, Comcast pays Level 3 far more to use their backbones than Level 3 pays to send data under this temporary agreement to Comcast's users. The net cash flow is towards Level 3, so they aren't really paying Comcast on a whole. That is, in my opinion, the real reason behind Level 3 agreeing to this until the FCC sorts it out. Level 3 could just stop serving Comcast and Comcast would be screwed but this would cost Level 3 more revenue than they saved by not paying Comcast these ridiculous fees.

Comcast lost revenue they had previously received for Netflix traffic and are attempting to be compensated. This is absurd and if it continues would set a bad precedent. Level 3 is still a Tier 1 provider. If someone has a different opinion, or I have my facts wrong please state it here instead of childishly editing the Wiki page repeatedly. —Preceding unsigned comment added by 146.6.93.87 (talk) 18:32, 1 December 2010 (UTC)

Agreed, making the change. Absolute Relativity (talk) 11:12, 2 December 2010 (UTC)

I have restored the content from before this whole issue started. This is not a peering agreement issue that effects status, and Comcast is not even a Tier 1 network in the position to affect Internet peering. Business disputes happen all the time; they are not encyclopedic material usually and this issue is not even settled yet. If ever, this may be of interest to network neutrality articles, not here. Kbrose (talk) 17:51, 4 December 2010 (UTC)

Actually this clear definition of "transit" makes everything about these definitions surrounding tiers much clearer. Even the article on Peering doesn't define it so well. A tier 1 network is then a network that is connected well enough that it has something to offer for any other network hence not being required to pay anyone. Any network in the periphery will have much reduced interconnection and has, typically, much more to gain by reaching everyone else than everyone else has by reaching it. This is basically what differentiates the tiers. It's basically about the size of the 'economy' you present and represent. However what interests me most is how the donut peering model influences the power of these 'core' networks in relation to the peripheral networks who can potentially create their own 'provincial roads' bypassing the 'central highways'. Xennex81 (talk) 14:09, 11 May 2014 (UTC)

Global Crossing

The article comments how only 6 of the tier 1 carriers are US based. I know Global Crossing is headquartered in Bermuda, but isn't this just for tax reasons? Do they actually have substantial operations there? AGeorgas (talk) 01:45, 7 December 2010 (UTC)

Data does not match source, AS Rankings

The refs for the tier 1 table are http://as-rank.caida.org/ and a diagram based on it. Why not just use the current numbers, rather than "September 2007"? The Sept. 2007 numbers are not even available form the sources, AFAIK. The closest is a diagram of July 2007: http://www.caida.org/research/topology/as_core_network/2007/images/ascore-simple.2007_big.png -Colfer2 (talk) 15:20, 13 December 2010 (UTC)

Is transit free important or low transit latency ??

I don't understand the fuss behind being a transit free network which means very little to actual users. However users do notice the difference in latency from one network to another. Hence shouldn't quality be a measure of latency rather than which network buys transit and which one is doing paid peering etc ??

Something like Knodes Index http://www.fixedorbit.com/metrics.htm or similar maybe a better measure for quality of a network. Comments ?? — Preceding unsigned comment added by A6v • 10:51, 10 January 2011 (UTC)

Useful link to the Knodes Index, probably belongs on the page in question. —Preceding unsigned comment added by 67.188.1.174 (talk) 06:50, 1 April 2011 (UTC)

Way too many people seem to thing that Teir 1 is something to look for in an ISP. In fact for single homed customers looking fo reliable service using a teir 1 provider is a BAD idea (and using a wannabe Teir 1 provider like cogent is an even worse idea) because it puts you one peering dispute away from losing access to parts of the internet. The Teir 1 providers provide an important role in terms of providing other providers with routes of last resort but generally a single homed customer who wants the best service should be looking for a Teir 2 provider with good links to at least two different Teir 1 proviers and preferably good local peering too. Plugwash (talk) 01:05, 6 November 2012 (UTC)

latest tier1 additions

"Telecom Italia, Deutsche Telekom and France Telecom (AS5511) all became transit-free during the year." http://www.renesys.com/blog/2011/01/a-bakers-dozen-2010-edition.shtml — Preceding unsigned comment added by 24.224.211.54 (talk) 16:32, 17 June 2011 (UTC)

France Telecom (AS5511) both listed as Tier 1 and Tier 2

The article lists AS5511 as Tier 1 then says it is actually Tier 2. This is contradictory and should be fixed. --a3_nm (talk) 23:22, 20 May 2012 (UTC)

This is fixed, it's transit free according to Renesys & CAIDA. (30 August 2012) — Preceding unsigned comment added by 2A01:7C00:100:1:BAD:F00D:DEAD:BEEF (talk) 16:07, 30 August 2012 (UTC)

How do you know if an ISP is Tier 1?

One of the major problems with this article is a lack of attribution. What third-party source can you go to to find out if an ISP has settlement-free peering and qualifies as Tier 1? The provider's website is not a good source. Renysys is good, but most of their information is behind a pay-wall. I would like to propose using a new tool from CAIDA, a non-profit at the University of California San Diego.

CAIDA AS RANK: http://as-rank.caida.org/

CAIDA AS RANK's Relationship Graph is an updated (last updated 19-Jan-2011...where is 2012 data?), easy-to-understand graphic showing a carrier's customers, peers, and providers. If a carrier does not have any providers (no red on the graph), I think they fit the definition of Tier 1. Here are some examples (takes a minute to load the graph):

• Global Crossing (only "provider" is Level3, who they are merging with)
• Verizon
• XO (full disclosure: I work at XO)
• Tata (not Tier 1)

...please add more.

This way, we can have informed, fact-based discussions about the list of Tier 1 networks. — Preceding unsigned comment added by Naterosenberg (talk • contribs) 00:19, 14 June 2012 (UTC)

"If a carrier does not have any providers (no red on the graph), I think they fit the definition of Tier 1." the definition of tier 1 given here is settlement free peering not mere peering. Unfortunately as others have said this is virtually impossible to determine from the outside.

Also that data source seems a little suspect to me. I wonder if some links are being misidentified. For example bytemark (a provider I use) claim to have 4 upstrams and presense at a number of peering points yet caidia claims they have 25 providers two peers and 7 AS customers. Even if we assume there are some errors in the information bytemark give about themselves having 25 providers seems unlikely to me. 130.88.108.187 (talk) 13:27, 18 June 2012 (UTC)

CAIDA's algorithm uses routing data to infer settlement agreements. That's how they are able to say who is a customer of whom. Red means a carrier is paying settlements to another carrier (not Tier 1) while blue means it receives settlements. I admit that there are no perfect ways to determine who has a network with settlement-free peering, but the article right now is just taking the carriers' word for it, and there is no attribution for some of the claimed settlements. If this article is going to have a list of Tier 1 networks, I think it's better to use data from an independent, non-profit (not supported by carriers) rather than taking the carriers' word for it. Naterosenberg (talk) 16:11, 22 June 2012 (UTC)

Cogent and AOL

Can someone verify Cogent and AOL (I guess ATDN?)? 2001:470:1F07:8C0:0:0:1:204 (talk) 03:32, 23 February 2014 (UTC)

Cogent listed twice

Cogent (AS174) is listed both in the Tier 1 list and the "Other major networks" list. Probably should be fixed. — Preceding unsigned comment added by 199.247.179.19 (talk) 19:03, 27 February 2014 (UTC)

Routing

The section on Routing seems to be quite inaccurate. The sources stated do not really verify the statements or assumptions. I'm going to try to rewrite it to the best of my ability... at this point. I have rewritten everything and expanded it greatly. Some of the logic is based on what you guys would probably call "original research". There are some PDF papers available online that provide extra intel but not much else. I have ventured into [3] to do a little exploration of public peering. In the process I came to understand "tier 1" and the whole tier model much better. I think I have provided both a logical and practical explanation of the tier 1 relationships. Let me know what you think if anything. I will perhaps do a rewrite of earlier sections as well to provide a much more consistent and clear definition.

Xennex81 (talk) 18:25, 11 May 2014 (UTC)

August 2014 update

Based on other entries on this discussion page and developments of the last two or three years, I've updated the list of Tier 1 networks. I've added citation for all carriers based on information from CAIDA (which seems objective) and ordered them alphabetically. Furthermore, I removed Cogent from the list of 'other major networks' (this was a duplicate) as they don't seem to buy transit and don't seem to have paid peering agreements (at least not on a verifiable basis), I removed AboveNet from the list of tier 1 networks as public information shows that they now have a paid peering agreement/buy transit from at least two tier 1 networks (Verizon and Level(3)).

If you feel this edit is erroneous, please correct me. Uhro87 (talk) 19:06, 6 August 2014 (UTC)

Graphic

The graphic to demonstrate the connection of different internet tiers has inconsistent language and randomly placed terms such as POP which is not defined on the page. It is more of a word cloud. Isn't it possible to place links inside of an svg graphic to at least link to the definition. Toyotabedzrock (talk) 02:45, 3 October 2014 (UTC)