Thoughts?
except that.... they're not the same. Both have completely different purposes. Lxc is a container, a local runtime, kvm is a complete environment that can run completely different architectures and environments. The performance of Lxc is actually surprisingly low, with only a 57% drop in latency over kvm, but then lxc (or lxd, same thing) is newer technology than kvm.
I like to use both, but not for the same purpose. I wish Lxc would evolve though, so that it would be more efficient. That would save it from becoming irrelevant once application virtualization breaks through, because that's a lot faster and versatile for the same application field as lxc in certain topologies.
Do you use Lxc for anything, or do you just set up a separate kvm container for everything?
Can someone fill me in on the significance of this? Wouldn't compressing the guests make them essentially slower? I'm a bit new to this, but from what I know KVM is still king when it comes to single containers.
It's more than LXC hence the different name which 'is aimed at giving you the full experience of virtual machines, the full security of a hypervisor, but much, much faster. Without all that virtualization overhead, you get the full underlying performance of your host environment. On bare metal, these containers are just as fast as the native OS on bare metal. In the cloud, you are getting subdivided machines without getting sub-par performance.'
(directed at everyone since the topic didn't have much information)
I managed to watch some of the video but not all of it. As far as I understand this tech (and Docker) the real advantages will be in combining containers with VM's and also being able to to snapshot and version control the containers.
One of the disadvantages (depending on your viewpoint of course) is that running containers on bare metal firstly means they do not have mobility to a new host and secondly that they are sharing the same kernel. From an availability and security perspective this is fine for development workloads but would not be good enough for production use. By placing containers into VM's the workload can be seperated and protected.
I may have got this wrong or it's something that was covered off later in the video, but in anycase I can really see contaners catching on with many companies that are really buying into the whole DevOps approach and managing infrastructure as software.