New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Undersea cables cut near Jeddah (Saudi), connectivity to/from Europe-MiddleEast-SouthAsia affected
The cable failures occurred near Jeddah, Saudi Arabia, and affected major systems, such as the South East Asia–Middle East–Western Europe 4 (SMW4) and the India-Middle East-Western Europe (IMEWE) cables. These cables are critical arteries for internet traffic connecting Europe, the Middle East, and Asia.
The South East Asia–Middle East–Western Europe 4 cable is run by Tata Communications, part of the Indian conglomerate. The India-Middle East-Western Europe cable is run by another consortium overseen by Alcatel-Lucent.
Some links here:
Personally, can confirm increased latency and packet loss to Europe from my probe in UAE.
Comments
Yes, no fun.
But to distinct the good from the bad:
@leapswitch immediately informed me by mail.
@ovh (Mumbai DC, Singapore, etc.) just sucks atm, esp. Mumbai (which btw always sucks networkwise, but now it sucks even more)
Host Universal (used by @ShockHosting) now uses PCCW for Asia<->Europe which - in comparison to e.g. any TATA line - is really stable atm while TATA is at ~ 25% packet loss and high latency.
That's why my internet speed has reduced quite significantly. Southeast Asia.
Takes kinda a detour but rerouting works fine.
Well hopefully it gets fixed soon, and not a deliberate attack. That is wild though, never thought of anchors being a possible reason they get cut or damaged.
ovh france servers latency increased from 120ms from my place in india to 450ms traffic routing through singapore
@diwakerd if latency is important for you:
Europe to Singapore Shockhosting is about 180ms atm via Host Universal via PCCW. OVH Europe to OVH Singapore 250ms to 350ms meanwhile (using TATA) ....
@Neoon would sure find a better route even.
But this still might be an idea to use a tiny proxy in Europe and one in Singapore to re-route traffic.
Singapore to Marseille is roughly 135ms via GSL, can't get any better than that. (or that's what I think
)
Interestingly when this happened, was not able to find any news about this, I thought something wrong with services I use.
How did you get 250-350 on OVH SG? Marseille to Singapore is mostly unaffected if the express routes (AAE-1 and Sea-ME-WE 5) are being used to bypass the hop at Middle East or India.
It seems to only affect routing over TATA and not other T1 providers.
If your provider has more upstreams than just TATA, you can ask them to reroute to those instead.
Indeed, pretty nice.
Interesting, for me it is routed via America lol.
TATA seems to be fine? Not sure.
Below is from Leaseweb SG.
Had this yesterday, it got better again. OVH GRA<->OVH SGP is now 150ms while e.g. OVH GRA<->Shockhosting SGP is now at 160ms.
But OVH GRA<->OVH BOM is still at >210ms as they re-route via SGP atm but much better as yesterday also. So they finally shuffled around their backbones at bit.
As @sh97 observed, even TATA seems to work again somewhat. But yesterday was not nice*.
(*still better than DTAG<->CF free tier)
This is just that specific far network due to Telia's shitty peering in APAC and Telia being the only T1 the far network is directly connected to/advertising this prefix over.
Have pushed it via Cogent, via the far network's T2 upstream. Will monitor.
FWIW:
I see significantly higher latency and lower download speed from @HostDZire's Mumbai DC to most of my european targets. Funnily though I also see a few targets with actually better results than usually (ie. before the current fiber problem), e.g. and particularly plusline.net (DE) which has over 100ms lower latency currently as well as about 50% higher DL speed!
Can confirm latency dropped to ~155ms.
Curious - do you have a direct link to Cogent, or is it via CDN77?
Direct + Telstra Global. Blend in SG/JP will likely change over the next 6 months as we move towards picking up more T1 capacity directly in these metros. Nothing locked in yet though.
Thanks, looks like this is incorrect then: https://bgp.tools/prefix/207.2.122.0/24#connectivity
Intentionally using Datapacket for Cogent inbound paths in SG, keeps the AS PATH length nice with the other T1s that see us as peering routes (via other T1s) so we don't drag traffic in over the Cogent direct path when it may not be the best. All prefixes are still advertised to Cogent directly for redundancy, the local pref is just dropped down.
Trying to avoid the exact issue the far network we were discussing before has.
FWIW:
The current state, as seen from Mumbai, didn't change (not significantly anyway). Most european targets show around 200 (tendency 250) ms to about 300 ms - now with 2 exceptions, both DE: plusline and core-backbone the latter even better (ca. 10%) than the former, both in latency and download speed.
Somewhat strange because AFAIK usually cables (from ASOC) land in FR, Marseille or in the UK. So, maybe in NL (which is very near DE), but nope, the NL results are very similar to the UK results and many others.
A look at the route shows that it seems to go via SGP, Reliance directly to DE, FRA core-backbone; latency is about 110 ms.
Then, for comparison, I checked the route to NL, AMS (@Clouvider and xtom). Those routes also go via Reliance and SGP but are over 100 ms (actually more like 150 ms) slower.
So it looks like most routes to Europe (from India), at least currently, go via SGP, but a few like e.g. core-backbone and plusline have found a way better route "around" the Saudi Arabia bottleneck.
No, I had a stable 130ms link to Mumbai from Europe.
Right now it sits at 174ms for me, was 182ms earlier today.
Bengaluru is lower than Mumbai right now, interestingly.
Congrats - wherever in Europe you tested ...
In case this is a pissing contest, I win because I see even better numbers than you (to at least two providers/DC's)
Nice.
Maybe because Bengaluru is closer to the main undersea cable (a side arm of which goes along the indian coasts). Just a guess, I don't pretend to know for sure.
I tested only Mumbai that is, from Mumbai to diverse european locations.
Marseille, not just kidding, central europe, Frankfurt.
Yes of course, depending from where you test, you get better results.
I just wanted to show, that the cable cut isn't as worse as it sounds.
Nah, most of the cables land in Mumbai.
Well, around 250 ms from Mumbai to most of Europe does look quite bad to me.
But again, my main point is how come that 2 european providers/networks are very significantly better/lower latency than the rest?
One suspicion I have is that they simply react quicker and quite likely have alternative routes ready "in the drawer" if needed.
Yes, but there also is ICE IV which lands in IN, Kochi and has a direct fiber from there across land to IN, Chennai which might well explain your a.m. observation (as Bengaluru is situated roughly on that axis) and also is significantly faster than going around the southern part of India possibly even via Sri Lanka.
That said, I still wonder how a cable problem in Jeddah can harm Europe - ASOC connectivity that much.
It just goes directly to India instead of going via SG,I just checked.
That could explain it indeed.
seems to be impacting UAE/india/TATA traffic only.
SMW4 is an old cable and most transit have redundancies on SMW5-AAE1, SG<>FR is relatively fine.
I can confirm it has gotten much worse in Mumbai. I have a home connection, and 2 servers in Mumbai and both are reporting significantly higher latency to my Netcup Vienna chicken. Normally the ping is 150ms and yesterday after this underseas cable news, it was 220ms+ and today it has gone worse to 300ms. It is now being routed via singapore to US, Santa Clara and then to Europe.
Edit: If I use Cloudflare WARP, the Cloudflare routing takes me through normal EU route but I'm still seeing elevated latency. Netcup's IPv6, however, is still showing 262ms ping.
Couple of good reads here
I think the situation is even worse in UAE, as compared to my indian chickens. Everything is super slow. DU (AS15802) routing has taken a big hit, and they're now passing a lot of traffic through China Mobile, taking more precedence over even IX traffic.
Etisalat UAE (AS8966) seems to have done a good job of working around this problem, they've got the latencies down by smarter BGP selections I guess.