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.
Splitting bandwidth between IP addresses
I was performing a speedtest-cli on a new dedicated server(not being used yet) and i notice that the download speed is capping at 3Gbit/sec and upload is at 1.8Gbit/sec on a 10Gbps port speed.
The reply of the technical support is as follows
we noticed that there is a /27 IP range configured on the 10G interface of the server, which means that the traffic/bandwidth/speed is divided among all the 29 usable IPs of the /27 IP range.
Isn't is supposed to be that every IP should be able to saturate the port provided other IPs are not using the server and that the host i am downloading from is capable of providing that kind of network?
Comments
I can not stop laughing at that support response.
Run away. Support is increasingly stupid and has no clue what they're talking about.
Edit: please name the provider with such monkeys on their support desk.
lol what even is that support response
I love how they say it's "divided among all the 29 usable IPs of the /27 IP range". Really? Including IPs that aren't even being used?
Most servers also have a /64 IPv6 range, which has 18,446,744,073,709,551,616 usable addresses in it. Going by that same logic, a single IPv6 address can only use 1/18446744073709551616 of the bandwidth, which for a 10Gbps port is around 1 byte every 5000 years.
Genius.
Sounds like a response from a cheap outsource team that really doesn't care. I really can't see a dc/server provider actually responding with that.
@WilliamProfuse
Can you take a look at that ticket as it seems its not going anywhere with the tech support
2775362
It gets better lol
Yeah, Psychz definitely has an outsource team doing low level support, the actual Psychz employees are always helpful though.. It's always been this way. I never have to ask for much except rdns changes now days.
I am hoping to get an actual Psychz employee , that's why i tag William
Well, it's unlikely they can help either for your "issue". You can't really expect to get 10Gbps on a speedtest.net server running speedtest-cli.. Run some iperf tests to some servers in the same region or something. You're not going to saturate a 10Gbps port on a speedtest server, the most I've gotten on the public servers is like 3Gbps (which is what you say you got.)
Psychz only provide a 200MB test file, let alone iperf test.Given the level of the support , it seems pointless to ask them to the iperf so maybe a psychz employee could help here
Where is your server, I could fire up a iperf in NYC to let you test just drop me a dm or even better Skype or discord me.
How about ram, are they divide
Might be worth trying another speedtest server,
Then use the ID value that comes up in the output,
I'm sure you're already aware there's zero point in trying speedtest-cli on a 10g server.
It's really more intended for residential connection testing, generally. Iperf or aria2c on a known unsaturated 10+g server is the way to test this.
oh yep, that's true for sure, I guess some (most?) speedtest servers are on 1Gbit aswell which could give a wrong idea as well
https://support.ookla.com/hc/en-us/articles/234578628-Speedtest-Server-Requirements
It's nearly impossible to saturate a 10Gbit/s port with a single-threaded download. And speedtest.net is vastly inaccurate when measuring big port speeds.
This summarizes my experience with Psychz fairly well. Outage after outage making simple network changes.