Ubuntu 22.04 Windows host, slow network

[tramp]

I am running ubuntu 22.04.3 on windows host with VirtualBox 7.0.

I have around 800Mbps download in windows, and around 70-80Mbps on ubuntu. I have tried increasing network buffer size using vboxmanage. I have tried changing adapters to bridge verse NAT and changing the adapter hardware in virtualbox. All of this does not help the network is painfully slow.

Is this a known bug in virtualbox?

[fth0]

This has been reported by a few users this year, but no-one was able to investigate the cause themselves, since it's time-consuming and needs some deeper networking knowledge in addition to the reproducibility.

FWIW, getting less than 50% of the available bandwidth is a clear sign of a problem, getting less than 75% is already debatable.

[mpack]

IMO, using unreliable timing methods inside a VM may also contribute to the perception.

[tramp]

I ran google's speedtest on both windows host and Ubuntu guest. Host 900Mpbs, Guest 90Mpbs.

I installed ubuntu on Windows Hyper V virtual machine system and ran speed test and got 900Mbps. So it is possible to have high speed networking. I do not like windows Hyper V as I can not redirect usb to the guest OS, which is something I use.

I also connected a USB N wireless adapter and redirected the USB to ubuntu and was only able to get 9Mbps on Ubuntu. So I am wondering if the whole layer between guest and host is slow as I was expecting much higher speeds, networking said 70Mbps connection so I was expecting at least 50Mbps.

I suspect there is a major flaw/bug in later versions of virtualbox as others have gotten higher speeds in the past.

This is on VirtualBox 7.0.10 with fresh install of ubuntu, so should be easy to replicate issue.

[scottgus1]

One thing I would recommend, based on Mpack's thoughts about not relying on VM-timer-based benchmarks, which can be unstable:

Find a large file to download that comes from a fast mirror, like a university-mirrored Linux ISO maybe, that will download close to your ISP's download speed on the host. Time the download with a stopwatch.

Then do the same download in the VM, timed with a stopwatch.

Please post the results here.

[tramp]

Why not use speedtest.net or http://speed.googlefiber.net/?

Note if the timing on guest OS is 10x off I think this would create a lot of other problems...

[scottgus1]

Actual download timings with a stopwatch are believed here. Benchmarks relying on virtualized timers are not.

Do as you will, and suffer as you do...

[tramp]

So you do not trust virtual timers created by virtualbox and expect them to have 10x difference?
If this is the case it sounds like virtualbox has bigger problems than slow network issues.

Additionally the only reason network speed was tested was that doing downloads was taking a long time. That is the problem of slow network, resulted in the testing of network speed. As such there are two sources of information pointing to slow network speeds. It has also apparently been noted as a problem by other users as stated above. As such more metrics showing the problem really exists appears to be redundant and waste of our time.

Also there is an easy test to confirm the problem exists. Take Windows 11, install latest virtualbox, then install Ubuntu 20.04.3, run any networking speed metric you want and then you can confirm or deny the issue.

Arguing over the metrics is waste of everyone's time, however if you want to do this then, "Do as you will, and suffer as you do..."

[scottgus1]

I refer you back to fth0's assessment:

This has been reported by a few users this year, but no-one was able to investigate the cause themselves, since it's time-consuming and needs some deeper networking knowledge in addition to the reproducibility.

Post a Bugtracker ticket with a log of the VM where the slowdown occurs.

BTW, we are not Oracle employees here. We are fellow unpaid volunteer Virtualbox users, like yourself. Since you don't have an Oracle support contract (you'd definitely know if you paid for one) you have come to us for free help. But you're not actually entitled to any help.

When you come to us to ask for help you are asking your equals, and you should be willing to carry a volunteer spirit to help us help you by giving us the data we feel we need and trust. When you push back and insist that you understand things better than us, despite your coming to us for help, that may move us to not want to help.

[tramp]

To that point my original question was:

Is this a known bug in virtualbox?

I will assume this is a known bug based on the responses, and as such I do not need to create a bugtracker. It also sounds like it is hard problem that no one is wanting to deal with or fix, oracle employee or otherwise, so will be in the product for a long time to come.

As for as Oracle support I would be happy to pay for that, but it sounds like I should investigate other VM solutions first and see if they are less buggy and if they have employees and developers monitoring forums. Also need to compare and contrast with just buying another PC, which is looking like the best path forward.

As far as fixing the code myself, I would consider doing this, but it appears the community is not very friendly and is argumentative about testing methods instead of dealing with the real issues. Actually this could be the source of the underlying problem and why no one wants to fix the bug.

[scottgus1]

You fought back on a simple and very polite request to provide actual download times. You've demonstrated the unfriendliness.

Locking to prevent further pointless discussion from a person who does not want to cooperate to help us help him/her solve the problem.