Tuesday, August 22, 2006

Containers vs. Xen: Everyone wins!

So containers are starting to get some press now. This seems to be a pretty fair issue. As I mentioned earlier in my blog, there is actually a mailing list where OpenVZ, vserver, and a number of community folks are actively engaged on bringing Containers into mainline Linux. Yes, the performance benefits of lightweight containers are great, the ability to do some resource tracking and throttling of containers will be useful for many end users and No, containers will not replace Xen. Xen provides a strong isolation model at increased overhead. Containers provide a low isolation model at decreased overhead.

Of course, with Containers, you have several options today to choose from, each of which have their drawbacks. vserver provides pretty complete isolation, has a great community supporting it, is available in some of the community supported distros, etc. There is also OpenVZ, supported by the SWsoft guys directly. And there are things like OpenMetaCluster from IBM, which can checkpoint/restart databases on clustered systems, although with a little more limited availability. All of those folks are working to get a common solution into mainline Linux, which should happen in time for the next big cycle of enterprise distros. Of course, in the meantime, the distros have each chosen a champion that they are considering. Since the changes for either solution are fairly invasive, they are outside the normally accepted changes for a service pack or update release. But in the meantime, let the fun continue!



At 12:22 PM, Blogger Steve said...

Yup i guess paravirtualization packages like Xen offer much more stability in terms of kernel crash problems over container based lightweight virtualization.What about Bochs and Qemu?What category do they fall into?


Post a Comment

Links to this post:

Create a Link

<< Home