ISP's with the 'best' (lowest) international latency

Forums › New Zealand Broadband › ISP's with the 'best' (lowest) international latency

1 | 2 | 3 | 4

Talkiet

4787 posts

Uber Geek

Trusted

#1985060 28-Mar-2018 17:28

ArcticSilver:

ctucks: Except I do notice a difference? I appreciate a lot of people that message because they don’t waltz in and immediately act like they own the place. I’m asking a question, and looking for constructive messages. It seems like you didn’t even read my initial post.

I feel like you’re the kind of guy that will tell people there isn’t a difference between 60hz and 120hz. People want to get the best available service they can, and I personally want to minimise my latency as much as possible.

Anyways, thanks to the people trying to help. I’m not trying to bring in any toxicity here, but it seems that every time I make a post, someone has to jump on their high horse and lecture people.

I think I’m gonna stick with my initial idea and get Orcon Fibre.

Instead of replying to everyone who is wrong (yes wrong) here, I thought id just add my support to you.

20ms of latency is DEFINITELY noticeable on time sensitive applications, so a 50ms+ difference to LA can be quite a jump with some games. Yes 20ms doesn't usually ruin a game, but it can have a bigger impact than people here seem to realize, but a lot of this is dependent on the games coding as well as the players ability to perceive the difference. 75+ ms is extremely noticeable and definitely NOT just in the users head.

I don't want to detract from the thread, but in short, some of the responses I found very arrogant and rude, at the end of the day if you don't have some thing constructive to say (or don't know what you're talking about) please don't post.

Hopefully Orcon works out.

You make valid and fair points, but I'm going to point out that some of the people replying here know more about these things than you could imagine, and have observed the impacts in individual tests, captures, troubleshooting as well as the impact over large userbases of latency changing. Some of us have written tests to explicitly modify latency millisecond by millisecond and then run tests and done packet captures on the results, AND THEN look at the throughput and behaviour of TCP over time to visualise the precise difference in TCP ramp rates as the latency increases from (say) 1ms up to 400ms.

You bet there are some arrogant responses here, and some are possibly rude, but there are a bunch that are very well informed and based on a lot of experience, domain specific knowledge,and relevant testing.

75ms+ increases will certainly have an impact to some applications/games... 20ms... Well it really depends what the base value was... 1ms up to 21ms, yeah, I'll give you that... 150ms up to 170ms. Nah... 30-50ms... Maybe, depends on the app.

Bear in mind as well of course there's 16ms (approx RTT) between Auckland and Chch, so the relative difference might be quite variable depending on where in NZ the user is.

Cheers - N

Please note all comments are from my own brain and don't necessarily represent the position or opinions of my employer, previous employers, colleagues, friends or pets.

Tradepub

Free professional, reference and technical white papers (affiliate link).

ArcticSilver

728 posts

Ultimate Geek

#1985107 28-Mar-2018 19:38

Talkiet:

You make valid and fair points, but I'm going to point out that some of the people replying here know more about these things than you could imagine, and have observed the impacts in individual tests, captures, troubleshooting as well as the impact over large userbases of latency changing. Some of us have written tests to explicitly modify latency millisecond by millisecond and then run tests and done packet captures on the results, AND THEN look at the throughput and behaviour of TCP over time to visualise the precise difference in TCP ramp rates as the latency increases from (say) 1ms up to 400ms.

You bet there are some arrogant responses here, and some are possibly rude, but there are a bunch that are very well informed and based on a lot of experience, domain specific knowledge,and relevant testing.

75ms+ increases will certainly have an impact to some applications/games... 20ms... Well it really depends what the base value was... 1ms up to 21ms, yeah, I'll give you that... 150ms up to 170ms. Nah... 30-50ms... Maybe, depends on the app.

Bear in mind as well of course there's 16ms (approx RTT) between Auckland and Chch, so the relative difference might be quite variable depending on where in NZ the user is.

Cheers - N

I feel its ironic that you say that people here know more about these things than "I can imagine", by saying that you are suggesting that I don't know what I am talking about.

I also find it interesting that you bring into this tests that are most likely related to TCP windowing which is mostly unrelated to response times in gaming and time sensitive applications, but lets not make any assumptions.

Thank you for the thoughts though.

hio77

'That VDSL Cat'
12999 posts

Uber Geek

ID Verified

Trusted

Lizard Networks

Subscriber

#1985111 28-Mar-2018 19:50

ArcticSilver:

I feel its ironic that you say that people here know more about these things than "I can imagine", by saying that you are suggesting that I don't know what I am talking about.

I also find it interesting that you bring into this tests that are most likely related to TCP windowing which is mostly unrelated to response times in gaming and time sensitive applications, but lets not make any assumptions.

Thank you for the thoughts though.

TCP windows comes down to a larger figure people care about, International speeds.

I have near identical Vodafone and Spark lines here. Spark line is literally light and day in terms of solid throughput, Vodafone is far easily affected and thus chainsaws.

Grain of salt given my line of business, however i work on the CX and process side rather than the core network. So to me, my major factor is... What do my customers care the most about?

The other thing to raw latency is, alternative paths can be taken or made to achieve a faster low end response time, but then you run into things like.... Does that lower path congest? is it fault tolerant? what is the cost of it?

Take for example the common game argued for latency, World of Tanks. hosted in SG it can take the route via US or directly from AU across to SG.

thing is, that far shorter route has pretty limited bandwidth, thus is not already the ideal option for all traffic.

#include <std_disclaimer>

Any comments made are personal opinion and do not reflect directly on the position my current or past employers may have.

ArcticSilver

728 posts

Ultimate Geek

#1985114 28-Mar-2018 20:03

hio77:

TCP windows comes down to a larger figure people care about, International speeds.

I have near identical Vodafone and Spark lines here. Spark line is literally light and day in terms of solid throughput, Vodafone is far easily affected and thus chainsaws.

Grain of salt given my line of business, however i work on the CX and process side rather than the core network. So to me, my major factor is... What do my customers care the most about?

The other thing to raw latency is, alternative paths can be taken or made to achieve a faster low end response time, but then you run into things like.... Does that lower path congest? is it fault tolerant? what is the cost of it?

Take for example the common game argued for latency, World of Tanks. hosted in SG it can take the route via US or directly from AU across to SG.

thing is, that far shorter route has pretty limited bandwidth, thus is not already the ideal option for all traffic.

I feel we're now talking two different topics.

Yes TCP windowing is important, for TCP connections, I'm not debating that, I'm just stating its not really relevant to the thread.

Again throughput is not really relevant (bar abnormally poor speeds) since this is regarding latency.

I agree that routing is very relevant and will change based on congestion and peering/interconnection arrangements and could change with any provider.

BarTender

3587 posts

Uber Geek

ID Verified

Trusted

Lifetime subscriber

#1985119 28-Mar-2018 20:21

ArcticSilver: I feel we're now talking two different topics.

Yes TCP windowing is important, for TCP connections, I'm not debating that, I'm just stating its not really relevant to the thread.

Again throughput is not really relevant (bar abnormally poor speeds) since this is regarding latency.

I agree that routing is very relevant and will change based on congestion and peering/interconnection arrangements and could change with any provider.

You would be surprised at how much difference policing and shaping has on the TCP window negotiated size makes in regards to your overall throughput especially when the window gets larger due to latency. The further away you get the worse poorly configured networks exhibit the problem.

You may say it's not important but if you tried playing with "tc" on a Linux based router and then seeing the real world impact it has you may change your opinion on if it's important or not.
And it's directly applicable to this thread.

ArcticSilver

728 posts

Ultimate Geek

#1985210 28-Mar-2018 23:49

BarTender:

You would be surprised at how much difference policing and shaping has on the TCP window negotiated size makes in regards to your overall throughput especially when the window gets larger due to latency. The further away you get the worse poorly configured networks exhibit the problem.

You may say it's not important but if you tried playing with "tc" on a Linux based router and then seeing the real world impact it has you may change your opinion on if it's important or not.
And it's directly applicable to this thread.

I'm confused, where did I say TCP windowing wasn't important? In fact, I believe i said the opposite. Throughput (by which I assume you are talking bandwidth?) is obviously important too if you want to do a large download or need multiple people to share a connection, but it is less important for gaming than latency (provided you have adequate bandwidth, which is extremely likely). One can impact the other, but again this thread is really related to UDP not TCP.

The point was that TCP isn't really that relevant for this thread as TCP is generally associated with applications that are not time sensitive such as downloading applications and loading websites.

Pumpedd

1759 posts

Uber Geek
Inactive user

#1985230 29-Mar-2018 08:08

I am on VF Cable and have been since forever. My latency to LA based servers used to be 140ish for many years and has been 170-200 for the last couple. It forced me to switch to Aus based servers that are around 40. If anyone tried to tell a gamer that any ping under 200 is fine, they clearly are not a gamer. When I play on US servers the difference is immediately obvious.

I did try discussing this with VF support but they just arent interested.

hio77

'That VDSL Cat'
12999 posts

Uber Geek

ID Verified

Trusted

Lizard Networks

Subscriber

#1985232 29-Mar-2018 08:12

ArcticSilver:

BarTender:

You would be surprised at how much difference policing and shaping has on the TCP window negotiated size makes in regards to your overall throughput especially when the window gets larger due to latency. The further away you get the worse poorly configured networks exhibit the problem.

You may say it's not important but if you tried playing with "tc" on a Linux based router and then seeing the real world impact it has you may change your opinion on if it's important or not.
And it's directly applicable to this thread.

I'm confused, where did I say TCP windowing wasn't important? In fact, I believe i said the opposite. Throughput (by which I assume you are talking bandwidth?) is obviously important too if you want to do a large download or need multiple people to share a connection, but it is less important for gaming than latency (provided you have adequate bandwidth, which is extremely likely). One can impact the other, but again this thread is really related to UDP not TCP.

The point was that TCP isn't really that relevant for this thread as TCP is generally associated with applications that are not time sensitive such as downloading applications and loading websites.

It's the balancing act where it becomes important.

You can't just say, "only gaming traffic go this way" without a substantial overhead compared to routing at the IP/BGP level.

While throughput and latency have the direct correlation in regards to window sizes, tcp acks etc you have to remember... They both go over the same network. They both are handled the same.

This is where "Gamer" plans could draw big improvements for pure latency. however generally those have not actually done this at a routing level as it becomes expensive.

Tools such as hast, killping do this at a much more granular controlled single user end, where the processing overhead is closer to neglectable.

#include <std_disclaimer>

Any comments made are personal opinion and do not reflect directly on the position my current or past employers may have.

BarTender

3587 posts

Uber Geek

ID Verified

Trusted

Lifetime subscriber

#1985336 29-Mar-2018 11:14

ArcticSilver: I'm confused, where did I say TCP windowing wasn't important? In fact, I believe i said the opposite. Throughput (by which I assume you are talking bandwidth?) is obviously important too if you want to do a large download or need multiple people to share a connection, but it is less important for gaming than latency (provided you have adequate bandwidth, which is extremely likely). One can impact the other, but again this thread is really related to UDP not TCP.

The point was that TCP isn't really that relevant for this thread as TCP is generally associated with applications that are not time sensitive such as downloading applications and loading websites.

But are you 100% sure that the service you plan to use is running over UDP rather than TCP. Many games still use TCP to communicate and to keep a persistent session open between hosts, it's not all UDP and you have zero control over if the game works over TCP or UDP as that is written by the game developer.

So TCP very much has relevance as a poor TCP throughput would result in overall poor experience.

If the app is pure UDP (unlikely) then it's a different story since then it would be up to the application to handle retries / lost packets. But the vast majority of traffic over the internet is TCP so how that works over large latency links is vitally important to your internet experience. Unfortunately you can't change the speed of light so latency is here to stay on our little pacific island.

1 | 2 | 3 | 4