ISP: Stem connect; issue with international latency

[chrisGolf]

Stem being your ISP, who is the underlying fibre provider. Openserve, Vumatel etc?

If Openserve for instance, you can trial out other ISPs to try narrow down where the issue lies.

As per your ref we don't use NTT, Orange would be our Upstream that would peer or buy capacity from NTT.

Vuma. It's not that easy to trial out other ISPs and even FNOs. we don't have another FNO in any case. I'm quite sure it's the ISP (not FNO in this case)

 

[chrisGolf]

Stem being your ISP, who is the underlying fibre provider. Openserve, Vumatel etc?

If Openserve for instance, you can trial out other ISPs to try narrow down where the issue lies.

As per your ref we don't use NTT, Orange would be our Upstream that would peer or buy capacity from NTT.

So, is it possible that STEM have not bought enough capacity on NTT - now I'm experience a kind of throttling because they routing me around about way?

 

[PBCool]

So, is it possible that STEM have not bought enough capacity on NTT - now I'm experience a kind of throttling because they routing me around about way?

Stem buys capacity from a transit provider, in this case, Workonline. So it's up to them to manage anything like that.

 

[chrisGolf]

Stem buys capacity from a transit provider, in this case, Workonline. So it's up to them to manage anything like that.

thx. would you say the routing via France and/or other surrounding countries is a possible indication of a capacity issue on STEMs side? i.e. why would the routing not be optimised/direct ?

 

[cavedog]

thx. would you say the routing via France and/or other surrounding countries is a possible indication of a capacity issue on STEMs side? i.e. why would the routing not be optimised/direct ?

I doubt but it might be the service stem has with work online. Maybe they primary EASSy only hence the latency.

 

[PBCool]

thx. would you say the routing via France and/or other surrounding countries is a possible indication of a capacity issue on STEMs side? i.e. why would the routing not be optimised/direct ?

Over multiple cable systems, those cables may be the most "direct," but due to the longer physical distance, you would have some increased latency.

 

[chrisGolf]

I doubt but it might be the service stem has with work online. Maybe they primary EASSy only hence the latency.

thanks for all the responses. this sounds plausible, but I guess there's no way to prove/workout what undersea cable is being used?
I'm guessing that WACS or other Westcoast cables are better (like Equiano) ...
I will ask them ...

 

[PBCool]

thanks for all the responses. this sounds plausible, but I guess there's no way to prove/workout what undersea cable is being used?
I'm guessing that WACS or other Westcoast cables are better (like Equiano) ...
I will ask them ...

Yeap, but typically, when buying a transit service, you can't dictate which cable system your transit provider will use to carry your traffic.

 

[chrisGolf]

Yeap, but typically, when buying a transit service, you can't dictate which cable system your transit provider will use to carry your traffic.

but EASSy to the UK is substantially slower? wouldn't your customers complain if some of them randomly started to get 200ms latency to the UK instead of the expected ~165ms ?

 

[PBCool]

but EASSy to the UK is substantially slower? wouldn't your customers complain if some of them randomly started to get 200ms latency to the UK instead of the expected ~165ms ?

It's not slower per se, just higher in latency.

But this is why we don't use East Coast cable systems as our primary routes.

 

[cpu.]

but EASSy to the UK is substantially slower? wouldn't your customers complain if some of them randomly started to get 200ms latency to the UK instead of the expected ~165ms ?

More than 90% of their clients don't care about latency. It's the gamers and techies that test and monitor it.

 

[cavedog]

but EASSy to the UK is substantially slower? wouldn't your customers complain if some of them randomly started to get 200ms latency to the UK instead of the expected ~165ms ?

Also traffic over EASSy to London isn't usually 200ms+ it's more like 175 to 181ms unless traffic bounces around a bit first like Marseille > Paris > Amsterdam then London.

 

[chrisGolf]

Also traffic over EASSy to London isn't usually 200ms+ it's more like 175 to 181ms unless traffic bounces around a bit first like Marseille > Paris > Amsterdam then London.

so this is exactly what i'm seeing. really frustrating. I', also feeling sorry for myself for using STEM - they are a percentage higher than your CISP/AFri/WebSquad ISPs in terms of monthly fees. I went with them hoping that their transit providers where good, and they had good capacity etc, only to find out that I'm getting horrible latency(200ms+). and their response so far is that they can only help me if I have a static IP - see response; (definitely not going to pay more for a static IP on top of an already high fee)

Please note the only solution is to use BGP communities to not advertise the prefixes to NTT or any of their other upstream. This option is only applicable/available to our business SLA paying clients that have static IP addresses allocated to them .

I will cancel with them - there is no reason to use them, their fees are higher, and their latency is higher. maybe for business accounts they are good? but their primary business does not seem to to include residential fibre?

I've given them a DDNS hostname - hoping that they can use this instead - else need to wait month and half to get out...

 

[PBCool]

Also traffic over EASSy to London isn't usually 200ms+ it's more like 175 to 181ms unless traffic bounces around a bit first like Marseille > Paris > Amsterdam then London.

Well, remember Eassy only runs from Mtunzini to Djibouti for most of SAs traffic. You then need another undersea cable like SEA-ME-WE or AAE-1.

Which would then land in Marseille anyway if the transit provider picks up transit from there, which most of them would do (or Paris). Versus an express route through Europe you will get inflated latency via Eassy to UK specifically, and slightly less to Europe equivalently.

 

[chrisGolf]

Well, remember Eassy only runs from Mtunzini to Djibouti for most of SAs traffic. You then need another undersea cable like SEA-ME-WE or AAE-1.

Which would then land in Marseille anyway if the transit provider picks up transit from there, which most of them would do (or Paris). Versus an express route through Europe you will get inflated latency via Eassy to UK specifically, and slightly less to Europe equivalently.

EASSy is a bad route to use in my opinion for UK traffic. If your ISP is pushing your UK traffic over this route I would avoid them?

another question - if EASSy was used - would I not pick it up in a MTR ? would there not be a hop/node for Djibouti or Sudan?

 

[PBCool]

View attachment 1750749

EASSy is a bad route to use in my opinion for UK traffic. If your ISP is pushing your UK traffic over this route I would avoid them?

another question - if EASSy was used - would I not pick it up in a MTR ? would there not be a hop/node for Djibouti or Sudan?

Nah as it is just layer2.

 

[chrisGolf]

So ironnicaly - the issues on EASSy coincide with an improvement on my UK traffic.

Another Internet cable fault hits South Africa

A fault on the EASSy cable has impacted Internet traffic since Tuesday.

mybroadband.co.za

Assuming they must have then used a better undersea cable to route traffic?
Still seem to be going via FRA ? so not exactly sure how to work out if WACS is being used or Equino etc?

Around the 20th Aug - latency went from around 200ms to a much better 166ms;



graph from the article;


Just to conclude this thread - STEM Connect not a great ISP - obviously not a popular ISP, so perhaps not worth even mentioning it. Also this is based off one users experience.

chris

 

[PBCool]

So ironnicaly - the issues on EASSy coincide with an improvement on my UK traffic.

Another Internet cable fault hits South Africa

A fault on the EASSy cable has impacted Internet traffic since Tuesday.

mybroadband.co.za

Assuming they must have then used a better undersea cable to route traffic?
Still seem to be going via FRA ? so not exactly sure how to work out if WACS is being used or Equino etc?

Around the 20th Aug - latency went from around 200ms to a much better 166ms;

View attachment 1752636

graph from the article;
View attachment 1752637

Just to conclude this thread - STEM Connect not a great ISP - obviously not a popular ISP, so perhaps not worth even mentioning it. Also this is based off one users experience.

chris

I don't think they are a "bad isp", I know the guys that run it.

When they select a transit provider, it's very difficult to ask your transit provider to steer traffic just for a few particular clients.

So for the most part they probably check the boxes for most, but if you need specific latency you need a specific ISP