dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
3040
dwd
join:2008-12-16
Eureka, CA

1 edit

dwd

Member

Latency from USA to Europe at New York.

Is there any workaround for this latency jump going from NY to Europe?

Host Name IP Address Hop Ping Time Ping Avg % Loss Pkts r/s Ping best/worst

173-219-251-220-link.sta.suddenlink.net 173.219.251.220 3 31ms
173-219-251-154-link.sta.suddenlink.net 173.219.251.154 4 23ms
xe-7-0-0.edge2.SanJose1.Level3.net 4.79.238.37 5 82ms
ae-2-70.edge2.SanJose3.Level3.net 4.69.152.81 6 85ms
sjo-bb1-link.telia.net 80.239.128.173 7 82ms

Here's where it starts:

nyk-bb1-link.telia.net 213.155.130.128 8 137ms
kbn-bb1-link.telia.net 213.155.131.134 9 176ms
s-bb1-link.telia.net 213.155.130.52 10 215ms
s-b1-link.telia.net 80.91.246.233 11 188ms
fre-peer3-link.telia.net 213.155.132.162 12 193ms
fre-c5-link.se.telia.net 81.228.75.100 13 192ms
fre-d1-link.se.telia.net 81.228.72.35 14 191ms
90-229-7-135.link.se.telia.net 90.229.7.135 15 185ms
62-20-84-14.customer.telia.com 62.20.84.14 16 188ms
* Unknown Host * 194.237.240.218 17 199ms
* Unknown Host * 194.237.240.242 18 195ms
jimbopalmer
Tsar of all the Rushers
join:2008-06-02
Greenwood, MS

jimbopalmer

Member

I think this 15 year old paper is still the clearest description of your problem, mostly starting in Section 2.

»www.stuartcheshire.org/p ··· est.html

It is a very good read. Thankfully dial up is not the huge bottleneck it was then.

SDL L3Tech
join:2011-06-07
Tyler, TX

SDL L3Tech to dwd

Member

to dwd
Due to MPLS, the actual jumps in latency can be misrepresented in a traceroute. However, if you directly ping the hops in question you'll see the true latency. Unfortunately, there is nothing I can do to reduce latency any further for traffic going from the US west coast, to the US east coast, then across the Atlantic ocean to Europe.

For example:

From Eureka CMTS to "sjo-bb1-link.telia.net 80.239.128.173 7 82ms" (San Jose)
ping 80.239.128.173

Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 80.239.128.173, timeout is 2 seconds:
!!!!!
Success rate is 100 percent (5/5), round-trip min/avg/max = 12/12/16 ms

From Eureka CMTS to nyk-bb1-link.telia.net 213.155.130.128 8 137ms" (New York)
ping 213.155.130.128

Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 213.155.130.128, timeout is 2 seconds:
!!!!!
Success rate is 100 percent (5/5), round-trip min/avg/max = 92/92/96 ms

From Eureka CMTS to "kbn-bb1-link.telia.net 213.155.131.134 9 176ms" (Somewhere in Europe, not sure what KBN stands for)
ping 213.155.131.134

Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 213.155.131.134, timeout is 2 seconds:
!!!!!
Success rate is 100 percent (5/5), round-trip min/avg/max = 180/180/180 ms

moldypickle
Premium Member
join:2009-01-04
Haughton, LA

moldypickle to dwd

Premium Member

to dwd
Considering you just bounced a signal there and back in 2/10ths of a second, I'm thinking the only thing you could do would be to move to the east coast...
dwd
join:2008-12-16
Eureka, CA

dwd

Member

Thanks. Move to the east coast, I know. Actually, now that I'm rereading this and looking at the tech's pings, it's about the same latency from CA to NY as it is from NY to SE, about 80-90ms. So add that up and you get 180ms give or take.

Since electronic signals can travel a near light speeds through copper cable, I'm wondering if we can get latency down t irrelevance?

For instance, Los Angeles to Stockholm is 5500 miles or 11000 miles round trip (ping) at 200ms. I think that's about 55, 000 mph or 29% the speed of light. That's really remarkable, really.

Could that number increase to 50% speed of light or are we up against physic's limitations already?

moldypickle
Premium Member
join:2009-01-04
Haughton, LA

moldypickle

Premium Member

keep in mind you are being routed several times also, you're not JUST crossing one line of copper.
dwd
join:2008-12-16
Eureka, CA

dwd

Member

Of course. That's why I'm asking if latency could be again dropped or halved? Perhaps something like dynamic routing where the origin and destination are taking into account, and routed accordingly, reducing hops to bare minimum.

SDL L3Tech
join:2011-06-07
Tyler, TX

1 edit

SDL L3Tech to dwd

Member

to dwd
Many things affect latency other than just the speed of light and distance. If the transceivers on the US side were powerful enough to shoot light across the ocean unassisted then the latency would likely be significantly lower. However, every connection, splice,repeater, router, and other devices can increase the latency.

Here is a good read in regards to transatlantic fiber:
»sites.google.com/site/bi ··· it-works

I guess you could think of the entire path as a long road with tons of yellow flashing traffic lights with a lower speed limit through those intersections.

moldypickle
Premium Member
join:2009-01-04
Haughton, LA
ARRIS SB8200
Ubiquiti UDM-Pro
Ubiquiti UniFi UAP-nanoHD

moldypickle to dwd

Premium Member

to dwd
said by dwd:

Of course. That's why I'm asking if latency could be again dropped or halved? Perhaps something like dynamic routing where the origin and destination are taking into account, and routed accordingly, reducing hops to bare minimum.

If every router along the way wasn't already doing this, you would probably just never reach your target, LOL.

Reducing hops to a bare minimum is only part of the issue also. If all connections did that, then you'd have some parts of the internet be utterly flooded. Also have to account for being routed around "traffic jams" or downed lines/hardware.

Also, WOW to the Emerald Express. Expected round trip NY>Eu>NY 62ms. Amazing
dwd
join:2008-12-16
Eureka, CA

dwd

Member

Please explain:

"Also, WOW to the Emerald Express. Expected round trip NY>Eu>NY 62ms. Amazing"

For the other who Posted. I understand, fundamentally, the problems with routing, traffic, down lines, power outs, etc.

I guess I was asking if it is a matter of economics or physical limitations? I'm sure that if we wanted to we could go from California to Stockholm digitally at near light speeds, for instance, but that would not be a practical application on the consumer market.

And if it is possible, how long till we see another drop in latency?

(I remember in 1998 and dial up that going 300 miles would produce 300ms+ latency, and not much better with cable.)

moldypickle
Premium Member
join:2009-01-04
Haughton, LA
ARRIS SB8200
Ubiquiti UDM-Pro
Ubiquiti UniFi UAP-nanoHD

moldypickle

Premium Member

If you read the link l3tech supplied, then you would see that Emerald Express is an upcoming transatlantic link. Upon further looking, that particular link is going to be the biggest laid yet and has an expected round trip latency from New York to Europe and back of 62 ms.
dwd
join:2008-12-16
Eureka, CA

dwd

Member

Unless I'm missing something, the Emerald link will not be any faster than I'm getting now:

From Eureka CMTS to nyk-bb1-link.telia.net 213.155.130.128 8 137ms" (New York)
ping 213.155.130.128

My ping from California to SE and back is around 180 to 200ms (see above).

I get to NY in 137ms (Actually I've gotten to a game server on the east coast of Canada averaging 80ms throughout a 2 hour gaming session).

But then my latency goes from 137 to average 190ms after NY. That means from NY to SE and back to NY is around 60ms lol.

What am I missing?

moldypickle
Premium Member
join:2009-01-04
Haughton, LA

moldypickle

Premium Member

Wasn't saying it was faster than your line.

Did you look up what the Emerald Express was going to be bringing to the table? If so, you would understand my more than geeky wow, if not, well, sorry.
dwd
join:2008-12-16
Eureka, CA

dwd

Member

From the link above:

Emerald Atlantis has planned a 100 Gbps undersea cable system, hopefully being implemented in late 2012. The Emerald Express is going to connect North America to Europe through Iceland as seen in Figure 7. This cable system has been designed to support 100 x 100 Gbps on each of its six fiber pairs. Emerald Express is projected to be eco-friendly; 100% carbon free and renewable energy powered data centers.(6)

Still not getting the WOW. Please explain what I'm not getting.

moldypickle
Premium Member
join:2009-01-04
Haughton, LA

moldypickle

Premium Member

Apparently I need to just go nerd rage my Trek dolls or something, LOL.

I enjoy reading and researching things such as this. That pipe is a really BIG and fast pipe they are laying. Very exciting to me.
dwd
join:2008-12-16
Eureka, CA

dwd

Member

It's only 100-GB each way. One of the pipes already there is over 1TB each way. That's what I wasn't understanding. Interesting to me too, but 100GB compared to 1TB+ is thing to get excited over.
jimbopalmer
Tsar of all the Rushers
join:2008-06-02
Greenwood, MS

jimbopalmer to dwd

Member

to dwd
Throughput is almost entirely unrelated to latency.

moldypickle
Premium Member
join:2009-01-04
Haughton, LA
ARRIS SB8200
Ubiquiti UDM-Pro
Ubiquiti UniFi UAP-nanoHD

moldypickle to dwd

Premium Member

to dwd
The 100Gbps pipe uses modulation..... as do they all. There is not just one single carrier on that fiber line.....

BASICALLY, the Emerald Express is being rated at a design speed of 4 pair x 100 waves x 100 Gbps = 40Tbps of throughput. If there is 6 lit pairs in the pipe and not four then it'd be 60 Tbps. (The actual company owning the cable says 4 pair but others say 6 pair, so not sure there) But 40 Tbps compared to 1 Tbps IS something to get excited over.

Also, the throughput are rated in BITS, not BYTES, big difference. Just so you don't get confused on it there.

jimbopalmer, throughput does become related to latency when demand > throughput. You'll either get forced idle or packet drops. But in reference to this conversation, I don't think either of us are saying that a bigger pipe is faster, it just happens that it is an immensely bigger pipe AND faster (or just as fast as other smaller ones).
dwd
join:2008-12-16
Eureka, CA

dwd to jimbopalmer

Member

to jimbopalmer
Jim, so true. I was thinking that at only 100Gb each way, latency may come as a bottleneck on that line.
dwd

1 edit

dwd to moldypickle

Member

to moldypickle
Or 42million Mbps.

ok my turn -WOW-.

That means 2.7 million people, at the same time, could download at 15Mbps and not saturate the line.

It does say 100x100, and I misread that at 100 each way.
the fastest pipe now is app. 11Tbps.

Still, if my latency is 100ms to NY and goes to 180 when jumping to Stockholm, I'm not going to see much difference, unless some of my latency is from saturation of current pipes.

I guess we'll just have to wait and see what happens.

Besides being "private" what is the difference between TAT and NonTat lines? Or in this sense, what does 'private' mean?

moldypickle
Premium Member
join:2009-01-04
Haughton, LA
ARRIS SB8200
Ubiquiti UDM-Pro
Ubiquiti UniFi UAP-nanoHD

moldypickle

Premium Member

Would have to do some more looking into it, but from the way I understand it is private lines are just that. I would assume that they either allow who they want on the lines or are purpose built for moving a shit ton of data for 'X' reason for 'X' company. TAT lines are built and owned by joint-ventures of different telecom companies. TAT/non-TAT is just 'public' vs 'private'. And when I say public, that might be misleading because you can't just jump on ANY tat line, you'll be getting routed across lines that particular provider is renting/owning.

But being this particular lines is also hitting Iceland and Netherlands i believe, there's a LOAD of chances that some very interesting services will be able to use it. Bottom line though, public or private doesn't really matter. The traffic is ALREADY on existing lines. More capacity across bottlenecks is only ever going to be a good thing!

Also, looking over some of the other older lines, best case latency on a majority of them is 100ms or better.
dwd
join:2008-12-16
Eureka, CA

dwd

Member

Well, the new line may not mean anything for us consumers then.
Moostang
join:2009-03-24
Tyler, TX

Moostang to dwd

Member

to dwd
If your latency is the same regardless of what time of day it is, it is not very likely that any of it is caused by saturation. Internet bandwidth has peaks and valleys and during the valley time latency should be unaffected.

moldypickle
Premium Member
join:2009-01-04
Haughton, LA

moldypickle

Premium Member

The internetz,,, it's made up of mountains and valleys.