Jailhouse: A Linux-based Partitioning Hypervisor
1–10 of 17 posts
Re: Jailhouse: A Linux-based Partitioning Hypervisor
#2It's a shame they haven't provided some benchmarks...
Re: Jailhouse: A Linux-based Partitioning Hypervisor
#3This is interesting, in that it seems to be in compeiting space to LXC.. but makes use of VMX for machine isolation rather than using namespace / cgroups. It's a shame they haven't provided some benchmarks...
I can see using this for real-time applications alongside management stuff and for separating critical and possibly buggy kernel drivers to where they can't harm the rest of the system.
Re: Jailhouse: A Linux-based Partitioning Hypervisor
#4This is interesting, in that it seems to be in compeiting space to LXC.. but makes use of VMX for machine isolation rather than using namespace / cgroups. It's a shame they haven't provided some benchmarks...
It can't be competing with LXC, with LXC you can do over-provisioning of the server and maximize the utilization. With Jailhouse you under-subscribe your system in order to get maximal separation and performance guarantees. I can see using this for real-time applications alongside management stuff and for separating critical and possibly buggy kernel drivers to where they can't harm the rest of the system.
I don't think the benchmarks make much sense in this situation, unless you measure server utilization and performance guarantees (which is the dimension in which they differentiate themselves).
Re: Jailhouse: A Linux-based Partitioning Hypervisor
#5Earlier quoted context omitted.
It can't be competing with LXC, with LXC you can do over-provisioning of the server and maximize the utilization. With Jailhouse you under-subscribe your system in order to get maximal separation and performance guarantees. I can see using this for real-time applications alongside management stuff and for separating critical and possibly buggy kernel drivers to where they can't harm the rest of the system.
Be fair. It can be said they compete, but, due to their very different design priorities, they don't compete directly. I don't think the benchmarks make much sense in this situation, unless you measure server utilization and performance guarantees (which is the dimension in which they differentiate themselves).
One could use LXC for completely allocating a CPU for some container so they can compete on one aspect. LXC still doesn't run on bare-metal and so can't take on the cpu separation for hardware accesses but there is a dimension in which they compete.
Re: Jailhouse: A Linux-based Partitioning Hypervisor
#6Re: Jailhouse: A Linux-based Partitioning Hypervisor
#7This is interesting, in that it seems to be in compeiting space to LXC.. but makes use of VMX for machine isolation rather than using namespace / cgroups. It's a shame they haven't provided some benchmarks...
Re: Jailhouse: A Linux-based Partitioning Hypervisor
#8For those who don't know, a FreeBSD Jail is a virtualization on the OS level (rather than Hardware level as provided by VirtualBox et al). It's similar in concept to creating a chroot, but you're also locking down processes.
Re: Jailhouse: A Linux-based Partitioning Hypervisor
#9This is interesting, in that it seems to be in compeiting space to LXC.. but makes use of VMX for machine isolation rather than using namespace / cgroups. It's a shame they haven't provided some benchmarks...
I don't know why you're comparing it with LXC, rather than e.g. Xen or KVM which seems much more directly comparable.
Re: Jailhouse: A Linux-based Partitioning Hypervisor
#10I've been working with FreeBSD Jails for quite a while now and this looks like a clone for Linux. For those who don't know, a FreeBSD Jail is a virtualization on the OS level (rather than Hardware level as provided by VirtualBox et al). It's similar in concept to creating a chroot, but you're also locking down processes. http://www.freebsd.org/doc/handbook/jails.html
Jailhouse appears to be almost like an exokernel.