Latency to overseas with fibre

[Deleted member 477823]

I wanted to know how good is the latency/ ping to servers overseas(most being EU) on a fibre connection. At the moment on my bad and ridiculously pathetic DSL connection my ping to a London server is 200+.

 

[Geoff.D]

I wanted to know how good is the latency/ ping to servers overseas(most being EU) on a fibre connection. At the moment on my bad and ridiculously pathetic DSL connection my ping to a London server is 200+.

That is almost as good as it will ever be. Where are you in SA?
Lets assume you are on WACS, which has a route length of 14000 km from the terminal in the Cape to the terminal in the UK.

A rule of thumb latency one way on fibre is 5 microseconds per km. That gives you a round trip time (RTT) minimum of 140 ms.

Now you have to add in the distance from you to the terminal in CT, plus the distance from the terminal in the UK to wherever the server is, plus all the equipment involved and you end up a rule of thumb of about 160 ms as the absolute minimum you can realistically expect. That is before any delays, queuing and other issues, before even considering the Bandwidth-delay product effects and the timing issues within your own PC and operating system.

 

[Johnatan56]

That is almost as good as it will ever be. Where are you in SA?
Lets assume you are on WACS, which has a route length of 14000 km from the terminal in the Cape to the terminal in the UK.

A rule of thumb latency one way on fibre is 5 microseconds per km. That gives you a round trip time (RTT) minimum of 140 ms.

Now you have to add in the distance from you to the terminal in CT, plus the distance from the terminal in the UK to wherever the server is, plus all the equipment involved and you end up a rule of thumb of about 160 ms as the absolute minimum you can realistically expect. That is before any delays, queuing and other issues, before even considering the Bandwidth-delay product effects and the timing issues within your own PC and operating system.

You mean millisecond I assume, 5 microseconds I'd be envious considering I am on fiber.

btw I get 146ms to London from CT, but agree with your 140ms for WACS.

 

[Geoff.D]

You mean millisecond I assume, 5 microseconds I'd be envious considering I am on fiber.

btw I get 146ms to London from CT, but agree with your 140ms for WACS.
View attachment 534981

No, do the calcs starting with the speed of light and the distances involved and the center of the band wavelength and the you get to the rule of thumb figure of 5 microseconds per km. Per kilometer. You can fine tune the figures if you want to be accurate by knowing what the quality of the fibre is and the refractive index of the specific fibre as well the specific wavelength used if you want. Over the band the values will vary between about 3.4 microseconds per km to about 6.6 microseconds per km.

The route length of the undersea cable to the UK on west coast is 14 000 km give or take a few km

Your figure of 146 ms is really good given you are in CT, and presumably fibre all the way.

PS
Just to P-off the proponents of fibre as apposed to radio wave in free space, the rule of thumb for radio is 4 microseconds per km. :crylaugh:

 

[Johnatan56]

No, do the calcs starting with the speed of light and the distances involved and the center of the band wavelength and the you get to the rule of thumb figure of 5 microseconds per km. Per kilometer. You can fine tune the figures if you want to be accurate by knowing what the quality of the fibre is and the refractive index of the specific fibre as well the specific wavelength used if you want. Over the band the values will vary between about 3.4 microseconds per km to about 6.6 microseconds per km.

The route length of the undersea cable to the UK on west costs is 14 000 km give or take a few km

Your figure of 146 ms is really good given you are in CT, and presumably fibre all the way.

I misread the original, thinking you mean 140 micro seconds when you wrote ms after writing out micro seconds instead of ms.

 

[Deleted member 477823]

That is almost as good as it will ever be. Where are you in SA?

Lol Joburg, that is obviously a huge distance from the Cape so it'll just add to the time xD

But anyway thanks for that, ja I'm just waiting to get fibre in my area but it seems as though it'll never come hence why I'm stuck with outdated broadband technology.

 

[Johnatan56]

PS
Just to P-off the proponents of fibre as apposed to radio wave in free space, the rule of thumb for radio is 4 microseconds per km. :crylaugh:

/takes bait
interference.

 

[Geoff.D]

Lol Joburg, that is obviously a huge distance from the Cape so it'll just add to the time xD

But anyway thanks for that, ja I'm just waiting to get fibre in my area but it seems as though it'll never come hence why I'm stuck with outdated broadband technology.

I have good and bad news for you. I am on a 2Mbps ADSL service in Pretoria. The Ping time to the same server used by Jonathan in his test is 215ms.

Assume we are both on Telkom's network all the way, then the difference is 70ms thereabouts for the extra distance between CT and PTA.
ADSL is just about all the way fibre already, except for the short last mile piece of copper as most ADSL SANS are already served by fibre. So, if you expect latency to improve drastically when you get fibre in your area, you are in for a disappointment.


AND BTW, the rule of thumb for a copper cable network EXCLUDING equipment effects is also 4 microseconds per km! YES Signals travel SLOWER though fibre than both radio and copper!

 

[Johnatan56]

I have good and bad news for you. I am on a 2Mbps ADSL service in Pretoria. The Ping time to the same server used by Jonathan in his test is 215ms.

Assume we are both on Telkom's network all the way, then the difference is 70ms thereabouts for the extra distance between CT and PTA.
ADSL is just about all the way fibre already, except for the short last mile piece of copper as most ADSL SANS are already served by fibre. So, if you expect latency to improve drastically when you get fibre in your area, you are in for a disappointment.


AND BTW, the rule of thumb for a copper cable network EXCLUDING equipment effects is also 4 microseconds per km! YES Signals travel SLOWER though fibre than both radio and copper!

Pinging Pretoria is 26ms. You should be getting around 180-190ms to London.


And signal traveling slower, you're forgetting overheads/interference/other factors which make it a worse medium for consistent latency.

 

[Geoff.D]

/takes bait
interference.

:crylaugh:

Ja when I gave my Sat comms course and dealt with latency etc it was hilarious to see the reactions when I tabled comparison figures between different media using the same end equipment and servers. It worked even better if I had access to a VSAT terminal, a dial up link, a ADSL link and a fibre link! Some were completely taken aback by just how much various OS's with their MTU settings has as well as the effects of the bandwidth-delay product had on throughput. It was vene more fun if I could show the difference between a Cisco router set with defaults and one optimised for the medium it was connected to!

No of course there are many things that affect performance, interference being one of them with radio. Unfortunately, these days most persons experience with radio is completely distorted by the way in which shared radio bands are used by WISPS over the shared 2.4 and 5 GHz bands. It is completely different story with properly controlled radio bands. In the former, interference is the single biggest issue, whereas in the latter, interference is managed by proper radio planning and sticking rigidly to the band plan.


So, most think radio ( microwave) is rubbish as a result, when it is actually the shockingly poor planning and deployment of the technology that is responsible for the so called poor performance many experience. And then I am assuming that there is no oversubscription going on in these deployments.

 

[The Voice]

Pinging Pretoria is 26ms. You should be getting around 180-190ms to London.
View attachment 534983

And signal traveling slower, you're forgetting overheads/interference/other factors which make it a worse medium for consistent latency.

And not taking into account the number of hops between your ISP and the endpoint in Europe.

 

[Geoff.D]

Pinging Pretoria is 26ms. You should be getting around 180-190ms to London.
View attachment 534983

And signal traveling slower, you're forgetting overheads/interference/other factors which make it a worse medium for consistent latency.

Yes agreed. For example, before Telkom deployed that stooopid ASSIA system of theirs, I consistently got ping time of about 25 ms to CT. After the ASSIA deployment that value shot up to 35 - 40 ms! so, I should subtract that effect from the difference to get the real added distance contribution.

 

[Geoff.D]

Ja, so we should ask OP to use pingplotter on his connection and then we can debate where all the latency is actually coming from!

 

[Geoff.D]

Pinging Pretoria is 26ms. You should be getting around 180-190ms to London.
View attachment 534983

And signal traveling slower, you're forgetting overheads/interference/other factors which make it a worse medium for consistent latency.

An extra 800 miles (1300km) should give 13ms for the distance component. So where does the 26ms come from? The difference is in the equipment and the routing. If one can have full control over the routing, one can get to about 15 - 20 ms between PTA and CT. That is EXCLUDING that ASSIA rubbish!

A few years ago I tested a 10Gbps FSO link between my home and the CSIR SANReN node. That gave me full control over the routing and I got to within 1ms of the theoretical RTT times between my home and a server in CT. 15ms theoretical and I measured 16ms.

 

[Deleted member 477823]

Ok so I got PingPlotter, what exactly is it that I should do in the program?

Also here's a quick speedtest I did...

Note that I'm hardwired and everything besides my PC is disconnected.

 

[juBa]

Ok so I got PingPlotter, what exactly is it that I should do in the program?

Also here's a quick speedtest I did...

Note that I'm hardwired and everything besides my PC is disconnected.

Had the same ping when I was on ADSL to EUW servers. I actually had 200ms but we are very far from the exchange and kept having disconnect/ reconnect issues on our line. Nobody could fix it, eventually they sent out a specialist and said that he’s putting us on a “different ADSL profile” that’ll increase ms but will be much more reliable. Ping jumped to 230ms but never had the disconnect issues again.

Then fibre came, 230ms shot down to 174ms in Pretoria. Never looked back since

 

[Auradion]

@ECProject97 Here - http://lg.jnb.psychz.net
use this server to ping international Servers
this as well. Ping yourself if you have Fibre or ping other Servers that are Gaming Servers and know the exact Latency

 

[ghostR]

On Cool Ideas from Cape town, you can expect about 141ms or so, Joburg you are looking at around 160ms +-

 

[Koosvanwyk]

Had the same ping when I was on ADSL to EUW servers. I actually had 200ms but we are very far from the exchange and kept having disconnect/ reconnect issues on our line. Nobody could fix it, eventually they sent out a specialist and said that he’s putting us on a “different ADSL profile” that’ll increase ms but will be much more reliable. Ping jumped to 230ms but never had the disconnect issues again.

Then fibre came, 230ms shot down to 174ms in Pretoria. Never looked back since

Here is on Fibre, from Cape Town, using Wifi

 

[Johnatan56]

Ok so I got PingPlotter, what exactly is it that I should do in the program?

Also here's a quick speedtest I did...

Note that I'm hardwired and everything besides my PC is disconnected.

Ping bras.afrihost.com
Ping speedtest.cpt.mweb.co.za
ping speedtest.jhb.mweb.co.za
ping vienna.at

First is exchange, second is CPT, third JHB, fourth Vienna (shows international).

Run each for ~2 minutes via pingplotter and check where the latency/jitter comes from.