Skip to content
Register

Contabo's "global anycast" is just two cities, same continent?

Networking by RAJ 24 replies 2.6K views
RAJ

Member OP

11 #1

Contabo is advertising "global anycast" on their new premium tier. I got two clients on it, recieved routing from both. Traceroute from my box in Dallas shows 12ms to their "anycast" IP. Same IP from my other box in... also Dallas (different DC). 11ms.

Their site says 15 PoPs worldwide. My clients want low latency for APAC users. If this is just two cities on the same continent, thats false advertising no? Any leads on actual global anycast VPS under $15/mo?

your margin is my opportunity
quinnbizz

Member

#2
RAJ said:
If this is just two cities on the same continent, thats false advertising no?

Take it with a grain of salt, but "global anycast" can mean different things to different providers. Some announce the same IP from multiple locations but only route efficiently from certain upstreams. YMMV heavily depending on your vantage point.

Have you tested from outside your own network? A single upstream or eyeball network can skew results. IMO worth checking from a residential connection or three before escalating.

...
adam20

Member

Ex-Abuse Desk

11 #3

The ticket said "your anycast is broken, my users in Indonesia see 180ms."

Turned out they were testing from a VPS behind the same transit provider who peered with us in only two locations. The user had never left the continent. We had to explain BGP communities to someone who thought "global" meant "works from my bedroom." I used https://bgp.tools to show them.

Dry humor aside, single-homed measurement is worthless. I've seen "global" anycast perform beautifully from 40 vantage points and terribly from one stubborn ASN.

reported. resolved. repeat.
SingaporeRep

Member

Benchmark Addict

#4 
$ ping -c 10 contabo.com
--- 10 packets transmitted, 10 received ---
rtt min/avg/max/mdev = 11.2/12.4/14.1/0.8 ms

$ ping -c 10 contabo.com (from HostHatch Tokyo)
--- 10 packets transmitted, 10 received ---
rtt min/avg/max/mdev = 178.2/180.5/195.3/4.2 ms

$ ping -c 10 contabo.com (from GreenCloudVPS Singapore)
--- 10 packets transmitted, 10 received ---
rtt min/avg/max/mdev = 176.8/179.1/182.4/2.1 ms

$ ping -c 10 contabo.com (from CloudCone Sydney)
--- 10 packets transmitted, 10 received ---
rtt min/avg/max/mdev = 185.3/188.7/192.1/2.8 ms

Decent disk IO, meh network

Comparison:
- Dallas: 12ms (works)
- Tokyo: 180ms (broken)
- Singapore: 179ms (broken)
- Sydney: 189ms (broken)

Single continent anycast, not global

fio, iperf, geekbench. results or gtfo.
nate_pad

Member

#5

If the IP is the same everywhere how do you know which server answers? Is there like a command for that? I only know traceroute sorry

JokoNord

Member

#6

@nate_pad, you can use looking glass from many provider, or RIPE Atlas probe. Very useful tool, I made VPS last month and learn much from forum here. Thank you very much

@RAJ bro, I test from Jakarta with my home connection and also from my VPS in RackNerd Los Angeles. Both show 170ms+ bro. Not good for my users. Contabo maybe need more PoP in Asia.

traffic worse than my packet loss
jane_ffm

Member

#7

We have active PoPs in six continents, not two cities

Single mode till I die 💀
MA

Member

#8

Single-homed traceroute is worthless.

garykwh

Member

Inference on a Budget

#9

170ms from singapore is rough tbh

CUDA cores are my love language
SingaporeRep

Member

Benchmark Addict

3 #10
jane_ffm said:
We have active PoPs in six continents, not two cities

Which six? Your site lists 15 PoPs but doesn't break down which tier has which. My tests above show Tokyo and Singapore both hitting 180ms. If there's a local PoP it's not being announced to the transit I can reach.

mark47 said:
Single-homed traceroute is worthless.

I ran four different upstreams. That's not single-homed.

fio, iperf, geekbench. results or gtfo.

Post a reply

You need an account to reply. Log in or register to join the conversation.

Post reply Preview Save draft