I saw this on an other discussion group and thought I would share. This is Cogent's financials because of the proposed ARC merger.


http://www.freeedgar.com/search/ViewFilings.asp?CIK=1158324&Directory=912057&Year=01&SECIndex=535608&Extension=.tst&PathFlag=0&TextFileSize=1905786&SFType=&SDFiled=&DateFiled=10/16/2001&SourcePage=FilingsResults&UseFrame=1&OEMSource=&FormType=S-4&CompanyName=COGENT+COMMUNICATIONS+GROUP+INC

Have fun!

Net income (loss) 2000 = $ (81,093 ), 2001 = $ (334,464 )
I wish I had that money to burn.
But - what's your point?

Cogent as a whole is losing alot more then that every month

Maybe you are missing "000" in those numbers.

Robert Marsh
Head Surfer Rackshack.net

Originally posted by Walter
Net income (loss) 2000 = $ (81,093 ), 2001 = $ (334,464 )


WOW!! $334 Million, and they want to acquire PSINet and its $3 Billion in debt??

Should be getting more and more interesting every day :)

But unfortunately, in this arena, Cogent is typical in many aspects. Lots of similar companies are losing huge amounts of money. The one advantage Cogent does have is they have a lot of money. I've given up predicting who the survivors will be. My only protective strategy is just not to be totally dependent on any one provider.

For instance, Level3-never has made money, and doesn't have a very large supply of cash...in process of raising more.

Verio for all practical purposes had to be purchased...and is still having to make major cutbacks.

I'm afraid we are dependent on many unprofitable companies at this time and have no idea where it all will lead.

For example, Amazon.com has never had a profit and, in fact, has lost more money each year they have been in existence. In the last year reported (2000), they lost $863 Million. How long can they do this?

Guess we just need to watch and stay prepared for whatever happens.

Originally posted by Incognito

For example, Amazon.com has never had a profit and, in fact, has lost more money each year they have been in existence. In the last year reported (2000), they lost $863 Million. How long can they do this?


Actually...they HAD never had a profit:

http://news.com.com/2100-1017-819688.html

Originally posted by uuallan


WOW!! $334 Million, and they want to acquire PSINet and its $3 Billion in debt??

That's actually a very smart move if they plan on filing for bankrupcy protection. I had a friend who's father lost his job and knew he was going to file for bankruptcy. So what did he do? He got as many credit cards and loans as he could and went on all kinds of vacations plus bought everything he could that could be traded for cash plus two cars and a house. So when he filed for bankrupcty, he got to keep one of the cars, the house, and all the things and cash he had gotten and then went back to school while living like someone who made 100K a year. Not really ethical but it sounds like what cogent might be doing.

So when he filed for bankrupcty, he got to keep one of the cars, the house, and all the things and cash he had gotten and then went back to school while living like someone who made 100K a year. Not sure what state he is in, but it really does not work that way anymore.

Originally posted by RackMy.com
Not sure what state he is in, but it really does not work that way anymore. I thought that Texas still has that law in their books where you don't lose your only house and your only horse. (or something like that) :)

At least they did thirty years ago.
I suppose things could have changed in thirty years. :D

Actually as far as large scale nationwide startups go, cogent didnt have alot of money, they got lucky and got in good with cisco, without cisco there would be no cogent.

It's easy.

*burp* Bonds, convertable debt, a lot of angry share holders, and shorts having a good ole time stickin it to a public Cogent. *burp*

These are the things that are in Cogent's future unless they find some miraculous way to make money. Amazon is making money now? Maybe they can stop floating all of those bonds to pay for their survival finally. I wonder.

Sounds like a whole mess of sour grapes to me. Lots of webhosting and colo providers locked into multiyear bandwidth agreements with the established Tier 1 gougers while startups aren't saddled with ancient pay per byte agreements.

Cogent's product and pricing is disruptive. Period. They are also good at delivering what they promise. Inexpensive bandwidth. I've been with them since April of last year and have no complaints. The price vs. performance is STELLAR. There's no way around that. No one else can touch them.

It's time to get out of those agreements you have with those buggy whip companies. There's a new sheriff in town.

http://www.lightreading.com/document.asp?doc_id=11087

[NG]Owner

It's time to get out of those agreements you have with those buggy whip companies. There's a new sheriff in town.Until they file for bankruptcy :)

Originally posted by [NG]0wner
Sounds like a whole mess of sour grapes to me. Lots of webhosting and colo providers locked into multiyear bandwidth agreements with the established Tier 1 gougers while startups aren't saddled with ancient pay per byte agreements.


Perhaps it sounds like a group of people who know the business very well, and have been around for a while, who are trying to understand a business model that has not shown it can make a profit yet?

I don't think anyone on this board will argue against the technical merits of the Optical Ethernet, and MAN technology. However, most of us are smart enough to spot another PSINet when we see one.

Originally posted by [NG]0wner
Cogent's product and pricing is disruptive. Period. They are also good at delivering what they promise. Inexpensive bandwidth. I've been with them since April of last year and have no complaints. The price vs. performance is STELLAR. There's no way around that. No one else can touch them.


The latest Boardwatch backbone performance numbers came out a couple of weeks ago. Cogent was dead last by a long shot.

That article about peering with PAIX - those are just public peering points. Established backbones have extensive private peering arrangements which keep traffic away from congested public peering.

Kevin

Cogent's product and pricing is disruptive. Period. They are also good at delivering what they promise. Inexpensive bandwidth. I've been with them since April of last year and have no complaints. The price vs. performance is STELLAR. There's no way around that. No one else can touch them.
But they aren't making any money, so how is their pricing relevent? I can sell you 1 dollar bills for 99 cents for a while, but sooner or later I'll be broke.

We also have had good experience with them - you can compare our cogent only trace server with some other sites out there - in fact I did traces last night and from europe and austraila - had significatly better traces than some other sites in the business.

they also have bent over backwards to fix even the smallest routing issue we find and are easily accessible.

they may not make it - but we are going to keep them in the mix until they go away. I like them - also it looks like in their financials they have at least another year of cash left.

you can trace to one of our cogent only sites at 209.51.128.22 to compare.

also - we have several of our customers' small linux boxes cranking out well over 5 meg each on cogetnt only - and I think they could do more. they are happy with them.

Use it while it lasts - but don;t bet the farm on it ;)

Originally posted by sailor
you can trace to one of our cogent only sites at 209.51.128.22 to compare.


Looks like they aren't peering with Cable & Wireless except at MAE-West, which means this trace goes from Pittsburgh to California before it even thinks about heading towards the final destination (Atlanta, Georgia, apparently).

3 bar2-serial5-1-1-0.Cleveland.cw.net (208.172.217.157) 21.141 ms 21.549 ms 20.950 ms
4 acr2-loopback.Cleveland.cw.net (208.172.210.62) 21.314 ms 21.569 ms 26.951 ms
5 acr2-loopback.SanFranciscosfd.cw.net (206.24.210.62) 83.173 ms 82.945 ms 83.191 ms
6 cpe1.MaeWest.cw.net (206.24.210.47) 78.260 ms 78.270 ms 77.756 ms
7 MAE-West.netrail.net (198.32.136.83) 65.903 ms 69.766 ms 64.623 ms
8 GigE2-1.tr1.SanFrancisco1.CA.us.netrail.net (205.215.12.1) 67.140 ms 67.914 ms 67.106 ms
9 sfo.netrail.demarc.cogentco.com (66.28.28.57) 80.923 ms * 81.338 ms
10 p15-0.core01.sfo01.atlas.cogentco.com (66.28.4.69) 84.384 ms 87.063 ms 84.879 ms
11 p4-0.core01.sjc01.atlas.cogentco.com (66.28.4.94) 82.957 ms 81.717 ms 85.999 ms
12 p13-0.core01.lax01.atlas.cogentco.com (66.28.4.73) 91.496 ms 95.115 ms 95.623 ms
13 p5-0.core01.san01.atlas.cogentco.com (66.28.4.78) 93.282 ms 94.142 ms 94.303 ms
14 p6-0.core01.iah01.atlas.cogentco.com (66.28.4.5) 90.932 ms 93.793 ms 94.884 ms
15 p15-0.core01.tpa01.atlas.cogentco.com (66.28.4.46) 93.906 ms 90.050 ms *
16 * p5-0.core01.mia01.atlas.cogentco.com (66.28.4.57) 93.671 ms 94.675 ms
17 p5-0.core01.atl01.atlas.cogentco.com (66.28.4.65) 91.736 ms 93.120 ms 94.830 ms
18 g49.ba01.b000173-0.atl01.atlas.cogentco.com (66.28.5.242) 95.660 ms 93.786 ms 92.537 ms
19 dv2.demarc.cogentco.com (66.28.28.254) 93.910 ms 90.603 ms 91.634 ms
20 dv2-atl-core-core-a.dv2.net (209.51.131.1) 94.396 ms 91.831 ms 96.411 ms
21 pluto.netdepot.com (209.51.128.10) 94.920 ms * 93.273 ms

Tracing to mindspring.com for comparison (they are also in Atlanta), it's nine hops and 25 ms.

Kevin

Here is one. A SD to SD route. Cari.net use to have the best connections in San Diego. Hops under 15ms point to point on Level3 and Genuity. They now push almost all traffic through Cogent. Our support office in SD to SDColo went from 8 hops to 21 using Cogent. Latency is terrible (mainly due to distance the route takes). Usually we are in the 200 and 300's, even though this one shows the best I have seen.

------------------------------------------------------------------
01/23/02 16:07:13 Fast traceroute SDColo.com
Trace SDColo.com (216.98.128.22) ...
1 64.160.0.0 20ms 30ms 20ms TTL: 0 (sndg02.pacbell.net ok)
2 63.200.206.152 30ms 20ms 20ms TTL: 0 (dist3-vlan50.sndg02.pbi.net ok)
3 206.13.30.6 30ms 20ms 20ms TTL: 0 (bb1-g6-0.sndg02.pbi.net ok)
4 144.223.30.201 30ms 20ms 20ms TTL: 0 (sl-gw22-ana-3-3.sprintlink.net ok)
5 144.232.1.41 30ms 20ms 20ms TTL: 0 (sl-bb20-ana-3-2.sprintlink.net ok)
6 144.232.18.241 60ms 50ms 50ms TTL: 0 (sl-bb23-fw-10-2.sprintlink.net ok)
7 144.232.8.110 80ms 80ms 80ms TTL: 0 (sl-bb26-pen-10-1.sprintlink.net ok)
8 144.232.16.89 90ms 80ms 80ms TTL: 0 (sl-bb25-pen-15-0.sprintlink.net ok)
9 144.232.5.38 90ms 80ms 80ms TTL: 0 (sl-gw4-pen-10-0.sprintlink.net ok)
10 160.81.202.18 90ms 80ms 81ms TTL: 0 (sl-cogentcomm-5-0.sprintlink.net ok)
11 66.28.4.1 90ms 90ms 80ms TTL: 0 (p5-0.core02.jfk02.atlas.cogentco.com bogus rDNS: host not found [authoritative])
12 66.28.4.86 110ms 141ms 110ms TTL: 0 (p14-0.core02.ord01.atlas.cogentco.com bogus rDNS: host not found [authoritative])
13 66.28.4.33 120ms 120ms 120ms TTL: 0 (p12-0.core01.mci01.atlas.cogentco.com bogus rDNS: host not found [authoritative])
14 66.28.4.37 140ms 130ms 130ms TTL: 0 (p5-0.core02.dfw01.atlas.cogentco.com bogus rDNS: host not found [authoritative])
15 66.28.4.25 140ms 130ms 130ms TTL: 0 (p15-0.core01.dfw01.atlas.cogentco.com bogus rDNS: host not found [authoritative])
16 66.28.4.98 130ms 140ms 130ms TTL: 0 (p13-0.core01.iah01.atlas.cogentco.com bogus rDNS: host not found [authoritative])
17 66.28.4.6 161ms 210ms 161ms TTL: 0 (p14-0.core01.san01.atlas.cogentco.com bogus rDNS: host not found [authoritative])
18 66.28.66.106 170ms 161ms 160ms TTL: 0 (g7.ba21.b006588-1.san01.atlas.cogentco.com bogus rDNS: host not found [authoritative])
19 209.126.155.112 170ms 161ms 160ms TTL: 0 (. bogus rDNS: host not found [authoritative])
20 209.126.155.111 101ms 210ms 100ms TTL: 0 (. bogus rDNS: host not found [authoritative])
21 216.98.128.22 90ms 90ms 90ms TTL:245 (www1.sdcolo.com ok)
------------------------------------------------------------------

I love Cogent for cheap bandwidth. However when you are paying full price for BW and they use almost all Cogent. It pisses you off.

Originally posted by UmBillyCord
Here is one. A SD to SD route.

...

I love Cogent for cheap bandwidth. However when you are paying full price for BW and they use almost all Cogent. It pisses you off.

Especially since your San Diego to San Diego route went by way of New York City! A mirrored analogue of my example, Pittsburgh to Atlanta by way of California.

Kevin

Originally posted by UmBillyCord
Here is one. A SD to SD route. Cari.net use to have the best connections in San Diego. Hops under 15ms point to point on Level3 and Genuity. They now push almost all traffic through Cogent. Our support office in SD to SDColo went from 8 hops to 21 using Cogent. Latency is terrible (mainly due to distance the route takes). Usually we are in the 200 and 300's, even though this one shows the best I have seen.

I love Cogent for cheap bandwidth. However when you are paying full price for BW and they use almost all Cogent. It pisses you off.

Yeah man that totally sucks how you got shafted like that. Letme know how the new place is you are moving your stuff to.

Yeah man that totally sucks how you got shafted like that. Letme know how the new place is you are moving your stuff to.

I'll keep you posted for sure. Today I signed the contract with you know who. I figure it will take about 6 months total to move everything. :D

I think I've got the fewest hops so far. I'm using Bellsouth.net DSL that rides out of UUNet and into Cogent. I'm in Gainesville, FL. It rides to Orlando, FL, then straight up to Atlanta, GA. 15 hops.

2 adsl-78-160-1.gnv.bellsouth.net (216.78.160.1) 11.731 ms 12.162 ms 12.818 ms
3 adsl-21-180-65.gnv.bellsouth.net (66.21.180.65) 13.220 ms 13.746 ms 12.556 ms
4 205.152.46.184 (205.152.46.184) 48.934 ms 237.079 ms 286.412 ms
5 205.152.46.37 (205.152.46.37) 67.222 ms 24.266 ms 20.949 ms
6 205.152.111.184 (205.152.111.184) 24.366 ms 22.054 ms 21.835 ms
7 500.POS4-1.GW5.ORL1.ALTER.NET (65.208.82.113) 18.681 ms 20.031 ms 18.620 ms
8 500.at-2-0-0.XR2.ATL1.ALTER.NET (152.63.80.110) 30.692 ms 30.791 ms 31.679 ms
9 194.ATM8-0-0.BR1.ATL1.ALTER.NET (146.188.232.53) 37.081 ms 32.982 ms 33.201 ms
10 t3-1-0-2.pr1.Atlanta1.GA.us.netrail.net (205.215.1.5) 47.080 ms 82.722 ms 109.094 ms
11 atl.netrail.demarc.cogentco.com (66.28.28.201) 52.365 ms 225.658 ms 122.628 ms
12 g49.ba01.b000173-0.atl01.atlas.cogentco.com (66.28.5.242) 179.867 ms 137.303 ms 54.135 ms
13 dv2.demarc.cogentco.com (66.28.28.254) 50.074 ms 135.598 ms 290.569 ms
14 dv2-atl-core-core-a.dv2.net (209.51.131.1) 55.734 ms 61.357 ms 60.540 ms
15 pluto.netdepot.com (209.51.128.10) 65.072 ms * 50.766 ms

I use NeoTrace to let me know where everything is travelling (it helps a lot in tracking fraud). Nice software.

Actually, I traced 209.51.128.22 and ended up with 209.51.128.10.

I think I got the longest one too ;)

traceroute to 209.51.128.22 (209.51.128.22), 30 hops max, 40 byte packets
1 66.129.95.2 (66.129.95.2) 0.436 ms 0.433 ms 0.359 ms
2 gige4-0.jax-core-02.peak-10.net (66.129.80.11) 0.349 ms 0.484 ms 0.314 ms
3 gige2-0.jax-core-01.peak-10.net (66.129.80.10) 0.360 ms 0.334 ms 0.360 ms
4 12.125.21.161 (12.125.21.161) 3.856 ms 3.799 ms 3.713 ms
5 gbr1-p51.ormfl.ip.att.net (12.123.32.138) 3.770 ms 3.756 ms 8.237 ms
6 gbr4-p70.ormfl.ip.att.net (12.122.5.125) 4.138 ms 4.266 ms 3.925 ms
7 gbr3-p20.attga.ip.att.net (12.122.2.181) 12.538 ms 13.602 ms 18.927 ms
8 gbr4-p60.attga.ip.att.net (12.122.1.142) 12.708 ms 14.015 ms 12.531 ms
9 gbr3-p20.dlstx.ip.att.net (12.122.2.89) 27.389 ms 26.755 ms 26.249 ms
10 gbr3-p30.la2ca.ip.att.net (12.122.3.70) 62.187 ms 65.089 ms 61.943 ms
11 gbr4-p20.sffca.ip.att.net (12.122.2.69) 73.467 ms 70.627 ms 70.296 ms
12 ggr1-p370.sffca.ip.att.net (12.123.13.69) 70.074 ms 70.795 ms 70.712 ms
13 nr1-p350.paix.ip.att.net (12.123.221.2) 77.535 ms 76.475 ms 76.374 ms
14 205.215.1.169 (205.215.1.169) 77.167 ms 76.636 ms 76.501 ms
15 GigE2-1.tr1.SanFrancisco1.CA.us.netrail.net (205.215.12.1) 76.608 ms 78.431 ms 85.649 ms
16 sfo.netrail.demarc.cogentco.com (66.28.28.57) 89.326 ms 89.725 ms 87.986 ms
17 p15-0.core01.sfo01.atlas.cogentco.com (66.28.4.69) 91.027 ms 90.121 ms 89.789 ms
18 p4-0.core01.sjc01.atlas.cogentco.com (66.28.4.94) 91.475 ms 92.279 ms 90.741 ms
19 p13-0.core01.lax01.atlas.cogentco.com (66.28.4.73) 98.671 ms 98.716 ms 101.257 ms
20 p5-0.core01.san01.atlas.cogentco.com (66.28.4.78) 157.852 ms 254.149 ms 207.645 ms
21 p6-0.core01.iah01.atlas.cogentco.com (66.28.4.5) 98.891 ms 99.639 ms 107.653 ms
22 p15-0.core01.tpa01.atlas.cogentco.com (66.28.4.46) 98.328 ms 97.781 ms 97.925 ms
23 p5-0.core01.mia01.atlas.cogentco.com (66.28.4.57) 112.149 ms 100.512 ms 97.647 ms
24 p5-0.core01.atl01.atlas.cogentco.com (66.28.4.65) 98.449 ms 99.022 ms 98.734 ms
25 g50.ba01.b000173-0.atl01.atlas.cogentco.com (66.28.5.246) 98.233 ms 99.204 ms 98.679 ms
26 dv2.demarc.cogentco.com (66.28.28.254) 98.922 ms 98.497 ms 97.856 ms
27 dv2-atl-core-core-a.dv2.net (209.51.131.1) 108.583 ms 105.654 ms 107.791 ms
28 pluto.netdepot.com (209.51.128.10) 119.521 ms * 99.344 ms

28 hops, oh my. This is from one of our servers in Jacksonville, FL. Bounced all the way to CA and back.

Last one I promise. From another datacenter we use in Gainesville, FL this time:

traceroute to 209.51.128.22 (209.51.128.22), 30 hops max, 40 byte packets
1 border1-fast0-0.Gainesville.atlantic.net (209.208.0.1) 0.571 ms 0.846 ms 0.462 ms
2 gsvlflma-br-1-s0-0.atlantic.net (209.208.6.125) 0.948 ms 0.86 ms 0.958 ms
3 209.208.90.29 (209.208.90.29) 9.028 ms 8.924 ms 9.112 ms
4 Serial2-11.GW7.ORL1.ALTER.NET (157.130.65.129) 9.144 ms 9.379 ms 10.012 ms
5 517.at-2-0-0.XR2.ATL1.ALTER.NET (152.63.84.214) 26.786 ms 20.679 ms 20.12 ms
6 194.ATM8-0-0.BR1.ATL1.ALTER.NET (146.188.232.53) 22.437 ms 20.012 ms 22.86 ms
7 t3-1-0-2.pr1.Atlanta1.GA.us.netrail.net (205.215.1.5) 40.263 ms 42.37 ms 40.101 ms
8 atl.netrail.demarc.cogentco.com (66.28.28.201) 39.66 ms 39.92 ms 40.454 ms
9 g49.ba01.b000173-0.atl01.atlas.cogentco.com (66.28.5.242) 40.881 ms 39.997 ms 40.278 ms
10 dv2.demarc.cogentco.com (66.28.28.254) 41.406 ms 39.795 ms 40.712 ms
11 dv2-atl-core-core-a.dv2.net (209.51.131.1) 44.031 ms 48.537 ms 47.614 ms
12 * pluto.netdepot.com (209.51.128.10) 44.891 ms *

Wow. Ok, no more. :)

I'm gonna jump back in here.

The reason I posted the link was this. One of the primary variable costs Cogent currently has (simplistic view of course) is transit links with Sprint and Above. The more customers they get, the higher those costs go. How does Cogent reduce that cost? Peering. Peering, aside from the capital expense to run the fiber optic lines into the PAIX facilities and the maintenance of the associated equipment, is FREE. By entering into that agreement with PAIX, and the subsequent agreements with others to exchange traffic there, Cogent won't pay Above and Sprint to carry that traffic, and Cogent will reduce its costs substantially.

Case in point: Why do you think Cogent bought Netrail? Peering. Why do you think Cogent is in talks to acquire PSI's US network? Peering. If you don't pay to exchange 90% of the traffic originating on or destined to your network, why won't Cogent's pricing work? Agreed, Cogent needs critical mass to make it work. It takes time (or acquisitions) to get there.

Also, it's counter-productive to compare the PAIX facilities to other public peering sites (MAE-West and East come to mind). Not all public exchange points are created equal. In general, are private peering points faster? Yes. Is the additional speed commensurate with the cost and effort required to secure dozens or hundreds of private peering agreements? Probably not.

I'm not sure I quite understand all the points of these posts traceroute posts.

Yes, Cogent has suboptimal traffic exchange with Cable & Wireless. Only one site on the West Coast. But that traffic exchange costs Cogent NOTHING.

Yes, Cogent has a problem with ATT. Only one site on the West Coast. But that traffic exchange costs Cogent NOTHING.

Does it take longer for the packets to get there? Yes. Do the packets get there? Yes. In 90% of web applications an additional 100ms doesn't make a hill of beans bit of difference. Given time, and additional peering, that 100 ms will disappear, and it will still cost Cogent NOTHING.

Regarding the boardwatch figures, Sigma, please post a link to that information. The latest Boardwatch figures I found regarding backbone performance was in March of last year. I'd be interested in seeing these more recent numbers.

I'll say it again. Cogent's pricing is disruptive. Cogent was able to start with a clean sheet of paper. Their network is robust and scalable. The key is this: If they can create a cost structure that is primarily fixed in nature (peering over transit), there is no reason to believe that their fixed price business model won't work.

As you can tell, I am a believer. I think Cogent has the right idea. And I think they can make it work. Are they deficit financing now? Yes. Will that turn around? I firmly believe so.

[NG]Owner

I truly hope the best for Cogent. I hope their debt doesn't overcome them like so many others.

Today: http://dailynews.yahoo.com/h/nm/20020128/tc/telecoms_globalcrossing_dc_5.html

Global Crossing -- Chapter 11. It's a shame.

Scott

Originally posted by [NG]0wner

Also, it's counter-productive to compare the PAIX facilities to other public peering sites (MAE-West and East come to mind). Not all public exchange points are created equal. In general, are private peering points faster? Yes. Is the additional speed commensurate with the cost and effort required to secure dozens or hundreds of private peering agreements? Probably not.


It is not counter-productive to compare PAIX with the MAEs. Carrier neutral peering points suck. If they didn't suck, then the tier 1/2 providers would not create private peering agreements with each other. They would just peer there.

You don't need 100s of peering agreements, you need multiple agreements with the top 10 backbones, because that is where 90+% of your traffic is going to go. The extra speed, and the reliability are most definitely worth it, otherwise the tier 1/2 providers would not do it.


Yes, Cogent has suboptimal traffic exchange with Cable & Wireless. Only one site on the West Coast. But that traffic exchange costs Cogent NOTHING.

Yes, Cogent has a problem with ATT. Only one site on the West Coast. But that traffic exchange costs Cogent NOTHING.

Does it take longer for the packets to get there? Yes. Do the packets get there? Yes. In 90% of web applications an additional 100ms doesn't make a hill of beans bit of difference. Given time, and additional peering, that 100 ms will disappear, and it will still cost Cogent NOTHING.


Tell that to fortune 1000 companies, tell that to sites like CNN and Amazon.com. Milliseconds matter. More importantly single points of failure ABSOLUTELY matter. A single connection, or even a single connection on each coast to major backbone providers is simply a way of setting yourself up for failure. Especially when so much of Cogent's traffic is destined for these major provider's backbones.

Also: additional peering will cost Cogent more money. How can it not? Eventually the public peering areas will be too restrictive and Cogent will be forced to enter private peering agreements, or suffer because their network lag times will get considerably worse.



I'll say it again. Cogent's pricing is disruptive. Cogent was able to start with a clean sheet of paper. Their network is robust and scalable. The key is this: If they can create a cost structure that is primarily fixed in nature (peering over transit), there is no reason to believe that their fixed price business model won't work.

As you can tell, I am a believer. I think Cogent has the right idea. And I think they can make it work. Are they deficit financing now? Yes. Will that turn around? I firmly believe so.



Level 3 started with a clean slate, they are so heavily laden with debt that people are starting to talk about bankruptcy. Cogent's network is only scalable internally. On the Internet, internal network scalability is only half the game. You have to be able to get traffic to other backbones with minimal latency. Without good peering agreements, they will not be able to scale their network, and they will either lose customers, or worse they will not be able to attract the kind of customers they need to prove they are a viable backbone option.

Deficit financing only works if you have another stream of revenue to cover the costs. UUNET built their backbone slowly, relying heavily on investment from Microsoft to build a dial network first, AT&T and Sprint both had/have phone revenues. PSINet had nothing else...and they went under. Aegis had nothing else, and they went under (as did Rhythms, Northpoint, etc). Cogent does not have another revenue stream that can off-set their network buildout, historically, this means they have a much better chance of failing than succeeding.

uuallan, as usual; Very well said!

Carrier neutral peering points suck. If they didn't suck, then the tier 1/2 providers would not create private peering agreements with each other. They would just peer there.

Circular logic. And again, I don't disagree that private peering is better. But I do believe that carrier neutral peering companies fulfill their obligations and provide a necessary and reasonable service to those who utilize them.

Tell that to fortune 1000 companies, tell that to sites like CNN and Amazon.com.

CNN and Amazon.com are not Cogent's target market. Fortune 1000 companies perhaps, but most likely only the bottom 600 of that list. I am assuming that most of these bottom 600 companies who would entertain using Cogent are looking for cost effective ISP to their employee's desktops, and perhaps to link their offices in major cities. For the latter, peering is irrelevent, it's all internal Cogent. For the former, milliseconds don't matter.

A single connection, or even a single connection on each coast to major backbone providers is simply a way of setting yourself up for failure.

A very good point. Which is why companies that select Cogent who are concerned with *any* single point of failure will have a backup connection(s) at the ready should something happen on Cogent's network.

Level 3 started with a clean slate, they are so heavily laden with debt that people are starting to talk about bankruptcy.

Most of Level3's debt is the result of putting fiber in the ground. Cogent has not put any substantial amounts of fiber in the ground, and certainly not between cities. Cogent bought existing dark fiber already in the ground at firesale prices (courtesy of the fiber glut precipitated by companies like Level3).

You have to be able to get traffic to other backbones with minimal latency.

This is a beauty in the eye of the beholder situation. For you, perhaps a 15-100 ms difference is worth the X thousand dollars more a month you will pay for a non-Cogent link of comparable capacity. For Cogent's target market, it isn't. The fact that Cogent doesnt' work for you, doesn't mean it won't work for anyone, and certainly doesn't mean that it won't work at all. For me it works, and it works handsomely.

Deficit financing only works if you have another stream of revenue to cover the costs.

Or until you reach critical mass.

Look, a Cogent connection may not work for you. But there are plenty of people out there who will take what Cogent offers and turn it into a business that works. The problem many others here are having is that the agreements they have with Sprint/Genuity/Level3/UUnet, etc., don't allow you the flexibility to competitively price your offering against those who do have Cogent. And that is problematic.

And for the majority of the people who are looking for websites or raw Internet access here don't need the redundancies and connectivity and rock bottom latencies that the price you are paying for those providers give them.

[NG]Owner

Originally posted by [NG]0wner


Or until you reach critical mass.

Look, a Cogent connection may not work for you. But there are plenty of people out there who will take what Cogent offers and turn it into a business that works. The problem many others here are having is that the agreements they have with Sprint/Genuity/Level3/UUnet, etc., don't allow you the flexibility to competitively price your offering against those who do have Cogent. And that is problematic.

And for the majority of the people who are looking for websites or raw Internet access here don't need the redundancies and connectivity and rock bottom latencies that the price you are paying for those providers give them.


The question implied in the original posting: Does Cogent have enough money to last until it reaches critical mass? Right now the smart money is on probably not.

If people can take what Cogent offers and turns it into a business that works, that's great, but what kind of business does not worry about the future? If you are building a web hosting business, and that's what we are talking about here, you have to start worrying about next year's bandwidth availability now. You also have to worry about whether or not your provider will be around next year, and if you will be able to keep your prices the same. If Cogent does not reach critical mass in the requisite time, they may have to double their prices to meet revenue goals. This means you may have to double your prices, and most customers will not like that -- after all, it won't be their fault you decided to sign up for Cogent. If I were running a hosting business, I would have no problem justifying my cost using Sprint versus Cogent, just show prospective customers who ask, the backbones and let them decide.

You haven't been in the hosting business very long have you? Everyone wants redundancy, uptime and low latency. Whether they are paying $3 a month or $300,000 a month. Read through the board and you will see people complaining about network performance all the time. When Cogent oversubscribes their network, and they will, the compaints will increase and hosts will have to dump them for a better provider.

I have a lot of respect for Charles from Affordable Colo, Mike from RackMy, and Frank from Hostopia, they have all put together companies that provide excellent value for the prices that they charge, and none of them use Cogent...there must be a reason for that.

Originally posted by [NG]0wner
Peering, aside from the capital expense to run the fiber optic lines into the PAIX facilities and the maintenance of the associated equipment, is FREE.

"Peering == Free" is by no means assured. Many providers do charge for peering arrangements, usually on a scale according to who sends more traffic to whom, and how much they can get away with charging. The deals are always locked under NDA, and if someone refuses to play along, there's usually a brief PR stink. In fact, PSInet had a disagreement with a few of its peers a while back, which makes it ironic that their network would be bought for its peering.

If you don't pay to exchange 90% of the traffic originating on or destined to your network, why won't Cogent's pricing work? Agreed, Cogent needs critical mass to make it work. It takes time (or acquisitions) to get there.


Well, in the ideal case you would have a huge transit-free network such as UUnet or Sprint, not pay for peering in most cases, and yet you would still have huge costs. The circuits themselves aren't too cheap.


Yes, Cogent has suboptimal traffic exchange with Cable & Wireless. Only one site on the West Coast. But that traffic exchange costs Cogent NOTHING.

Yes, Cogent has a problem with ATT. Only one site on the West Coast. But that traffic exchange costs Cogent NOTHING.

Does it take longer for the packets to get there? Yes. Do the packets get there? Yes. In 90% of web applications an additional 100ms doesn't make a hill of beans bit of difference. Given time, and additional peering, that 100 ms will disappear, and it will still cost Cogent NOTHING.


This is analogous to: Is my Yugo slower than a Honda? Yes. Does it still get there? Yes, unless there's a failure. Does it matter if I drive from Philadelphia to Chicago by way of San Antonio? No, as long as I get there.

If you're saying that Cogent is the Yugo of network providers, that makes sense. But you said it, not me :)


Regarding the boardwatch figures, Sigma, please post a link to that information. The latest Boardwatch figures I found regarding backbone performance was in March of last year. I'd be interested in seeing these more recent numbers.


Boardwatch has not put the content online. I don't know if/when they will do so. It's in the "Buyer's Guide & Backbone Directory 2002", and it arrived in the mail just a couple of weeks ago. The performance measurements were done by Keynote, which is a well-known comprehensive testing system, utilizing their fifty-three domestic test stations. They didn't do any tests to reflect quality of overseas connectivity.

Cogent's mean performance was slightly better than Touch America and ICG, but the variability was the highest within that group, causing them to be listed last. They were one of three providers with an "error rate" of 1.0 percent or above; the others were AboveNet and Aleron.

Unfortunately, many backbones did not participate in the performance testing, such as Global Crossing, Intermedia, etc. And the Sprint numbers were much lower than I would expect.


I'll say it again. Cogent's pricing is disruptive. Cogent was able to start with a clean sheet of paper. Their network is robust and scalable. The key is this: If they can create a cost structure that is primarily fixed in nature (peering over transit), there is no reason to believe that their fixed price business model won't work.


I wish them only the best luck.

Kevin

Originally posted by [NG]0wner
Peering ... is FREE.

Yes, Cogent has suboptimal traffic exchange...

Does it take longer for the packets to get there? Yes. Do the packets get there? Yes. In 90% of web applications an additional 100ms doesn't make a hill of beans bit of difference. ...

Regarding the boardwatch figures, Sigma, please post a link to that information. ...


I wrote a longer post but my browser crashed. I'll sum up.

- Peering is not necessarily free. Many providers require settlement fees under NDA. Some providers play big gorilla to try to force other networks to agree to fees. PSInet was a big offender in this area, back in the day.

- Your argument about packets going cross-country twice is analogous to this: Will my Yugo get me there? Yes, unless it breaks down. Will my Yugo get me there quickly? No. But hey, it's cheap. If you want to say that Cogent is the Yugo of network providers, then OK, but you said it, not me :)

- The Boardwatch numbers aren't online; they are in the "Buyer's Guide & Backbone Directory 2002" which came in the mail a couple of weeks ago. The performance testing was done by Keynote Systems from their fifty-three US testing stations.

Kevin