“I’ve always been fascinated by the idea of running an operating system under another operating system. It’s a cool hack to make a piece of hardware do double (or triple, or quadruple…) duty by running virtual machines. I’ve been hearing such good things about User-mode Linux (UML) that I had to give it a whirl.” Read the article at UnixReview.
UML is a great way to do that without having to use an unstable kernel for your workstation. There are a number of other testing scenarios where it’s very …
Oh don’t leave me hanging!
I checked the HTML, and it really does just end mid-sentence. Am I missing a link somewhere? I’m using Mozilla.
What’s the bottom line?
how does this compare with FreeBSD jail?
Interesting article. I don’t see much use for UML for myself, but I can certainly imagine situations where it would be quite useful.
I don’t think he gave an estimate of the overhead involved with running a UML virtual machine, which is something I’m interested in.
-Erwos
I did some extensive research about a year ago comparing the two for a developer level virtual hosting company I started. The main difference between the two is that UML runs another kernel, whereas a FreeBSD jail uses the same kernel with some restrictions placed on child processes.
I found it easier to set up a UML “distribution” once I got the userland kernel compiled correctly (I used gentoo). The FreeBSD jail was more error-prone to set up (more than once i did make installworld into the wrong location).
The downside to UML is that memory allocation has to go through two kernels. Another downside is the forced use of a “loopback” filesystem. This also adds overhead. In fairness, I ended up doing the same thing with the FreeBSD jail.
The downside to FreeBSD jails are that you can only have one IP address per jail, where UML allows any number. Setting up the virtual switch was harder in UML because of this.
In the end I chose FreeBSD because it offered a more stable hosting environment. I would choose UML if I needed something for personal use.