Features

Mouse pointer integration also worked out of the box on Windows. All you need to do is move the pointer into the virtual machine to enable the mouse and keyboard on the machine. Dynamic resizing of the guest desktop to fit the virtual machine window size only works if you have the extensions in place. This also gives you a full-screen mode and a seamless mode, which switches the desktop to full-screen mode, but you only see the Windows taskbar and the programs launched under Windows (Figure 4). Both modes worked without any trouble on Windows in our lab. Seamless mode also launched on Ubuntu but hid the Unity program launcher, making it impossible to use Ubuntu.

Figure 4: An unusual sight: Internet Explorer on an Ubuntu desktop. Seamless mode worked without any trouble in our lab.

To access the virtual machine remotely, you need to Enable Server in the Remote Display tab of the Display section. I had no trouble accessing the system with the KDE KRDC tool. VirtualBox uses RDP on port 3389 as a protocol; just make sure you enter the password for the host system when prompted, not the guest system.

File Exchange

Sharing the clipboard worked without trouble when copying snippets of text or URLs between the guest and host. However, dragging and dropping files between the systems only worked between the host and the guest. Where this did not work, installing the guest extensions typically resolved the problem; this step is often indispensable after a Windows update. Dragging and dropping between the guest and host did not work at any time in our lab environment. We only saw a small window with a prompt telling us to enter a file name.

On Ubuntu, drag and drop did not work at all. Of course, the desktop environment could be responsible for this. In a parallel test with Arch Linux as the host and Windows 7 as the guest, file exchanges worked without any trouble in both directions. And on Ubuntu 15.04 as the guest, at least you can transfer files to Ubuntu, although trying to send them back to the host caused the file manager to crash.

If you want to share folders on the host system with the guest, you need to go to Shared Folders in the settings and select the desired directory. At setup time, you decide whether VirtualBox makes the directory read-only and automatically mounts it. The software then sets up the appropriate SMB share, to which Windows directly assigns a drive letter (Figure 5).

Figure 5: Windows integrates shared folders into Explorer without trouble.

This method of sharing files worked on the fly in our lab without any problems. In the virtual machine menu, choose Devices | Shared Folders | Shared Folders Settings. On Ubuntu, I was unable to see either the network computers or the shares in Nautilus. The only thing that helped here was a detour to the command line, where I entered

$ sudo mount -t vboxsf <share> \
  <mountpoint>

for at least a modicum of success: Ubuntu mounted the share read-only, although this was not what I wanted.

Speed

In the performance test, I wanted to see whether the systems on the virtual machines were lame ducks when virtualized or fast movers. The first test was the transfer rate of a USB stick. Writing to the host system achieved an average speed of 11Mbps, whereas the Windows guest took approximately twice as long – achieving a speed of around 5.5Mbps for the same operation.

The situation was even worse for file reads: The host system achieved a speed of 22.5Mbps on average, and the Windows guest crept along at just 6Mbps. Ubuntu impressed in this exercise with an average read speed of 15.5Mbps and an average write speed of 10Mbps.

The situation was different in the Peacekeeper benchmark, which I ran on the latest Google Chrome version on all the systems. Windows surprised here by achieving a score of 4500 and thus actually beating the host system. Ubuntu was the laggard here with a score of just 2500.

VirtualBox shone in the network throughput test, with both Windows and Ubuntu guests; the values were more or less identical to those of the host system. The number of connections per minute differed greatly, however. Whereas the host set the bar at 2,000 connections per minute, the guests achieved less than half this figure. Of course, this is not really important in practice and should be viewed more as an academic value. It seems that Oracle has done a good job of implementing the network stack in the system.