Replies: 2 comments 1 reply
-
After running Iperf3 tests on my IONOS VPS server, I was getting 1.60Gbit coming direct into the server so unlikely to be VPS bandwidth, I also phoned them up to check the speeds and they said it should be 1Gbit. I've run Iperf3 tests on my home network between client devices and the OMR and they are all running correct at 1Gbit. That only leaves either my WAN interfaces or something in the server/ OMR config that I'm missing. I'm wondering if because I have 3 5G modems all in the same place, the radio tower is throttling the sim cards to split the total available bandwidth. If I remove one of the modems, the bandwidth of the other 2 does increase in NETWORK>>MPTCP>>BANDWIDTH and the overall total stays close to 300Mbit so perhaps I'm simply at the limits of what my local cell tower can provide? I think it is fed via Microwave link so might not be a full speed tower. This would be easy for me to confirm as I have a Starlink system also (currently offline). Next time I pay for a months usage on this I may try and connect it up as an extra interface and see if I can get better speed. There are other 5G providers in my area but the signal quality is poor for most of them, so don't think I could just buy sim cards from other providers and expect better speed. |
Beta Was this translation helpful? Give feedback.
-
After reading your initial post, my question was going to be if the 5G was the same provider. It sounds like it is. And also to speedtest your VPS with iperf or speedtest-cli. But, for a personal comparison, I've seen well past 500mbps with LTE/Starlink/WISP on my setup and OMR has been tested into the 10gbps+ range. My best guess here is you're hitting the limit of the cell tower, especially if it's backhauled by microwave. Even though theoretical 5G speeds are higher, actual throughput also has to do with the carrier backhaul and provisioning of the actual equipment. More modems doesn't necessarily mean higher speed, carriers typically split bandwidth (basically, by dividing airtime) among the connected devices. Most major cell carriers also have advanced quality of service solutions that allow them to control the bandwidth from devices, entire accounts and even specific protocols or destination network providers. This is done to ensure network stability, so clever people like you can't overwhelm the system. While a secondary LTE carrier is likely to help, signal strength is definitely part of the throughput calculation. But, my real question is that if you have fiber, aren't there simply better plans you could buy? |
Beta Was this translation helpful? Give feedback.
-
So far I have tested 3 VPS providers, Ionos, Kamatera and Lightsail.
All VPS have 2 Vcores and 1GB Ram. The server CPU usage max is around 40%. Ram usage is low less than 500MB.
I'm using Xray Vless.
My interfaces are 1 x 72Mbit Fibre Link. 3 x 5G link. The 5G links can hit around 200Mbit each max, but fluctuate between 40-200Mbit depending on time of day.
Each link has its own dedicated 1GB physical ethernet port on my router which is an Intel Xeon Dell R730 server running proxmox. The Lan connection also has its own dedicated port. All ports are running in VirtIO mode.
I've tried allocating various core counts, 2 cores and the CPU usage hits 100% (with SQM on) and my limit is around 250Mbit. I've given it 6 cores for some headroom although it also works ok on 4.
The Lan connection also has its own dedicated port. All ports are running in VirtIO mode.
I have tried with both SQM turned on and off.
I can't seem to break through the 300Mbit speed barrier. I've tried changing lots of settings in the router such as the proxy method, scheduler, congestion control etc but unable to get past this.
Ionos quotes that their VPS have 1Gb uplink. However I've found conflicting information about this here https://lowendtalk.com/discussion/169939/ionos-vps-s-benchmark-results where the VPS connection has been quoted at only 400Mbps which matches what I'm seeing with the MPTCP overhead.
The Kamatera VPS quotes a 10Gbit server connection but does not give any further information on actual internet uplink speed assigned to VPS.
Lightsail is a bit confusing as they quote a 2TB transfer, but I think this might be the traffic allowance not the speed. I checked on VPSbenchmarks.com though and download speed tested was 433Mbit, which with the MPTCP overhead would probably get me back to the 300Mbit I have.
There are minor speed differences between the 3 VPS, lightsail is a bit slower than the other two. Lightsail is probably the weakest VPS though as you aren't allowed to use anymore than 10% CPU or it comes out of your "bursting allowance". Once the bursting allowance is depleted, you are capped at 10% CPU which limits the transfer speed to around 100Mbit.
All 3 are quite cheap though at around £2-£7 per month. I've put this on hold over Christmas so as not to mess up the family internet.
So I'm thinking that this 300Mbit is as fast as I am going to get unless I can find another VPS provider. I'm going to do some more research on VPSbenchmarks as only came across this website today and possibly trial some other ones. Is it possible that multiple VPS providers cap the bandwidth at 400Mbps? Has anyone managed to exceed this figure?
Beta Was this translation helpful? Give feedback.
All reactions