NY ranks #1 in Internet b/w

Mailman List Mirror (Read Only)
1

Perhaps Sept 11 was no coincidence and perhaps they knew of these numbers as well. -Hank

http://www.totaltele.com/view.asp?ArticleID=45236&pub=tt&categoryid=626

New York is the Internet capital - TeleGeography
By Total Telecom staff

30 October 2001

New York is the world's Internet capital in terms of bandwidth connectivity, according to Washington DC-based research group TeleGeography Inc.

With almost 150 Gbps of region-to-region bandwidth connectivity, the Big Apple is hooked up to more than two-thirds of all interregional Internet capacity worldwide. London comes a long way second, with just over 85 Gbps.

According to the TeleGeography report, Packet Geography 2002, New York has direct connections into 71 other countries, 10 more than London.

The report ranks global Internet cities according to their roles as "interregional hub cities," measuring how much Internet capacity links them to other world regions. After NY and London, Amsterdam, Paris, and San Francisco complete the Internet global city top five.

In all, five of the top 10 cities are in the U.S., four are in Europe, and one, Tokyo, is in Asia. Although Europe and Asia each have major hub locations, most Internet traffic between Asia and Europe still traverses U.S. coast-to-coast routes. Indeed, according to the report, 13 of the top 25 companies providing international Internet connections in the U.S. are based outside of North America.

The Top 10 in full is:

1. New York (149,989.5 Mbps of Internet bandwidth)
2. London (85,518.7 Mbps)
3. Amsterdam (24,479.6 Mbps)
4. Paris (22,551.8 Mbps)
5. San Francisco (20,813.6 Mbps)
6. Tokyo (16,745.5 Mbps)
7. Washington DC (13,261.2 Mbps)
8. Miami (11,912.4 Mbps)
9. Los Angeles (11,227.0 Mbps)
10. Copenhagen (10,417.0 Mbps)
Note: Figures represent Internet bandwidth connected across international borders to Consolidated Metropolitan Statistical Areas or equivalents. Domestic routes are omitted. Data as of mid-2001.

This report can be purchased through Total Telecom's 'Reports and Research' resource. Click here for more details.

2

But if you look at trunks going into *another* country the same report comes
to this ranking.

London
Paris
New York
Amsterdam
Frankfurt

This report also says that the relevance of US for Internet is decreasing.
As ever: never trust a statistic unless you faked it yourself ...

-- Arnold

3

But if you look at trunks going into *another* country the same report comes
to this ranking.

London
Paris
New York
Amsterdam
Frankfurt

This report also says that the relevance of US for Internet is decreasing.

Hm, I'm still waiting to witness a traceroute from Europe to Asia or the
Pacific that doesn't go over the US for the first time. Are there subs
that can lay undersea cables yet? A cable from Northern Europe to Japan
and the US North West under the North Pole icecap would be great.

As ever: never trust a statistic unless you faked it yourself ...

But one thing is obvious: we IP people put our stuff where we think we
want it, not where it should go looking from a redundancy/vulnerability
standpoint.

If I want to send a packet from The Hague to Philadelphia, the packet will
almost certainly pass Amsterdam and New York, two places where huge
amounts of traffic can easily be disrupted. If the IP routers were to be
placed closer to the places where seacables surface, this problem would go
away: all those major hubs are serviced by multiple fiber landing
locations.

4

that can lay undersea cables yet? A cable from Northern Europe to Japan
and the US North West under the North Pole icecap would be great.

Don't know about the icecaps -- the Middle East will probably have more demand than the North Pole for a long long time -- but I think FLAG has some bandwidth they'd like to sell you (http://www.flagtelecom.com/cable_route.htm). As might the telcos who own SeaMeWe-3 (Disponible – Nom de domaine en vente)…

So:

> This report also says that the relevance of US for Internet is decreasing.

Hm, I'm still waiting to witness a traceroute from Europe to Asia or the
Pacific that doesn't go over the US for the first time.

Yes, but at least you can see traces from S Korea to Japan, say, which don't route through Palo Alto. Region-to-region is harder.

cheers
Bram

5

Hm, I'm still waiting to witness a traceroute from Europe to Asia or the
Pacific that doesn't go over the US for the first time. ...

You will find that there are many routes from Europe to Asia (and back again) that don't run via the US. Two examples...

  http://www.bbeng.gxn.net/cgi-bin/lg.pl?query=trace&addr=www.hkt.net&router=Amsterdam1

  http://www.bbeng.gxn.net/cgi-bin/lg.pl?query=trace&addr=www.singnet.com.sg&router=Amsterdam1

I'm sure there are many more. There are a few cables that run between Europe and Asia (SMW3 and Flag) that are good communications routes for non-US-transiting traffic.

Martin

6

>that can lay undersea cables yet? A cable from Northern Europe to Japan
>and the US North West under the North Pole icecap would be great.

Don't know about the icecaps -- the Middle East will probably have
more demand than the North Pole for a long long time -- but I think
FLAG has some bandwidth they'd like to sell you
(http://www.flagtelecom.com/cable_route.htm). As might the telcos
who own SeaMeWe-3 (Disponible – Nom de domaine en vente)…

Obviously the big advantage of the Mediterranean/Red Sea/Indian Ocean
route is that there are many countries along the way that can benefit from
the cable. But if you want to go from London to Tokyo, it's not quite the
shortest path: the great circle distance is 9600 km, but the FLAG is
nearly three times that at 28000 km. That's even nearly 10000 km more than
a London - New York - San Francisco - Tokyo route. The icecap route
would be only 12000 km.

>Hm, I'm still waiting to witness a traceroute from Europe to Asia or the
>Pacific that doesn't go over the US for the first time.

Yes, but at least you can see traces from S Korea to Japan, say,
which don't route through Palo Alto. Region-to-region is harder.

Right. The problem is that it's hard to justify expensive connections to
neighboring countries when:

1. You hardly exchange any traffic with them anyway (for instance, in The
Netherlands it's about 45% domestic, 45% to/over the US and only 10% of
the traffic goes to other European countries)

2. You need (and therefore have) a big pipe to the US anyway that can
absorb this traffic easily

It is extremely annoying when doing the right thing is too expensive.

7

Apparently, there's no enough EU <-> AP traffic to justify direct
circuits. The dispersion-shifted single-mode ground fiber (along the
route of Trans-Siberian railroad) does exist.

--vadim

8

Apparently, there's no enough EU <-> AP traffic to justify direct
circuits. The dispersion-shifted single-mode ground fiber (along the
route of Trans-Siberian railroad) does exist.

as does flag

randy

9

[snip]

Hm, I'm still waiting to witness a traceroute from Europe to Asia or the
Pacific that doesn't go over the US for the first time.

[snip]

Enjoy:

10

Rafi Sadowsky wrote:

>
>
[snip]
>
> Hm, I'm still waiting to witness a traceroute from Europe to Asia or the
> Pacific that doesn't go over the US for the first time.

Well, back in 1995 I connected a Japanese internet customer (the name
escapes me) to BTnet in the UK. The connection method was frame relay
and so at the IP layer there was one hop between London and Tokyo....

Nigel

11

Hm, I'm still waiting to witness a traceroute from Europe to Asia or the
Pacific that doesn't go over the US for the first time.

and we should not forget carl malamud's ds3 around the world in the
mid-90s. it was called park.org. not sure where to find info about
it on the net today.

randy

12

I don't know about AP/EU traffic, but I know that this trans-russion link is
already saturated (they are thinking abiut moving to WDM).

13

> Hm, I'm still waiting to witness a traceroute from Europe to Asia or the
> Pacific that doesn't go over the US for the first time.

Enjoy:

Thank you. :slight_smile:

10 london-bar2.ja.net (146.97.35.6) 12.378 ms 11.973 ms 12.099 ms
11 ten155-gw.ja.net (193.63.94.15) 12.781 ms 12.940 ms 12.753 ms
12 janet.uk.ten-155.net (212.1.192.149) 14.474 ms 18.835 ms 16.487 ms
13 212.1.192.186 (212.1.192.186) 303.626 ms 301.981 ms 303.517 ms
14 nacsis-2-FE2-0-0.sinet.ad.jp (150.99.99.17) 317.214 ms 313.987 ms
313.982 ms

Unfortunately, in this instance avoiding the US doesn't seem to pay off:

5 so-1-1-0.TR1.AMS2.Alter.Net (146.188.8.86) 5.187 ms so-1-1-0.TR2.AMS2.Alter.Net (146.188.8.90) 309.476 ms 296.644 ms
6 so-2-0-0.IR2.DCA4.Alter.Net (146.188.11.222) 87.299 ms 86.886 ms 86.995 ms
[...]
11 POS6-0.BR3.DCA6.ALTER.NET (152.63.38.117) 87.579 ms 86.699 ms 86.749 ms
12 a3-0.uunet.mclnva02.us.bb.verio.net (204.255.169.90) 320.341 ms 289.424 ms 291.342 ms
13 p16-0-0-0.r02.mclnva02.us.bb.verio.net (129.250.5.254) 291.258 ms 88.383 ms 88.220 ms
14 p4-6-1-0.r00.plalca01.us.bb.verio.net (129.250.2.245) 173.876 ms 173.900 ms 403.277 ms
[...]
18 p1-1.nii.snjsca03.us.bb.verio.net (129.250.10.22) 595.339 ms 567.663 ms *
19 kyoto-10-GE2-0.sinet.ad.jp (150.99.163.65) 444.986 ms 397.353 ms 336.185 ms

(I removed 4 UUNET routers in DC and 3 Verio routers in the Bay Area.)

Over 300 ms for less than 10000 km (6000 miles) is not great. Even a good
satellite should be able to provide better round trip times... But then
the GSM cellular network provides a minimum RTT of 550 ms even for very
short distances, so I suppose it could be worse.

14

Over 300 ms for less than 10000 km (6000 miles) is not great. Even a good
satellite should be able to provide better round trip times...

That sounds somewhat erroneous. Geosynchronus orbit is about 22,500 miles;
up+down+roundtrip makes that 22,500 * 4, or 90,000 miles;
90,000 / 186,000 miles/sec = 483 milliseconds, which, or course, due to
routers inducing very measureable delay, and the fact that an IP Packet
takes adds a little delay due to its lenght, is usually a bit more.

When working on this stuff (specifically on a hop from Oslo to Jerusalem),
I recall 630 ms being the average latency.

But, to make my point, Geosync orbit could never, ever be less than 483
milliseconds, ever.

-- Alex Rubenstein, AR97, K2AHR, alex@nac.net, latency, Al Reuben --
-- Net Access Corporation, 800-NET-ME-36, http://www.nac.net --

15

> Over 300 ms for less than 10000 km (6000 miles) is not great. Even a good
> satellite should be able to provide better round trip times...

That sounds somewhat erroneous. Geosynchronus orbit is about 22,500 miles;
up+down+roundtrip makes that 22,500 * 4, or 90,000 miles;
90,000 / 186,000 miles/sec = 483 milliseconds, which, or course, due to
routers inducing very measureable delay, and the fact that an IP Packet
takes adds a little delay due to its lenght, is usually a bit more.

I'm afraid you're right. 240 ms is firmly burned into my mind, but this is
the ONE WAY delay.

But, to make my point, Geosync orbit could never, ever be less than 483
milliseconds, ever.

Well, actually there is a caveat: the distance to the satellite is never
exactly 22500 miles. Depending on whether the orbit is measured from the
surface of the earth (which is obviously the case for regular
non-geosynchronous satellites) or the center of the earth (which I think
is done with the 22500 mi figure) the satellite is either farther away or
closer, depending on the location of the observer and the orbit of the
satellite. The difference is substantial: up to 4000 miles.

16

Practically, I know back in 1994 when I was in the US and the Nordic
university network had a backup E3 satellite link I had to use, the round
trip times were approx 600ms. 483ms round trip if you're at the equator,
sounds like it can correlate fairly well to 600ms when you have to go
from northern europe, to a satellite hanging up above the atlantic ocean,
and then down to north america (or might even be so satellite over africa,
then to satellite hanging over central america, then down to north
america, ie two satellite hops).

17

Few folks I guess recall the dial-up line that ran westward from Beijing to
Karlsruhe that functioned in the early-to-mid '80s. It ran at 300 baud. On
good days. -s

18

steve wolff wrote:

Few folks I guess recall the dial-up line that ran westward from Beijing to
Karlsruhe that functioned in the early-to-mid '80s. It ran at 300 baud. On
good days. -s

Steve, sure I do and Arnold Nipper on the list will as well since he I believe
implemented it... Was it not 9.6? Later is X.25........

Dave

19

That was an X.25 line, and I am not sure that it used IP atop the X.25. --SteveG

20

Steve Wolff wrote:

Few folks I guess recall the dial-up line that ran westward from Beijing

to

Karlsruhe that functioned in the early-to-mid '80s. It ran at 300 baud.

On

good days. -s

It was an X.25 dial-up running at 9.6 kbit/s and was turned down 94/95 when
IP connectivity to CN was well established.

-- Arnold (former co-worker of Prof. Zorn)