LXD, the Hypervisor for LXC Containers, Part Two

Last time, ContainerJournal corralled comments from Dustin Kirkland, Strategist, Container Offerings, Canonical on the LXD hypervisor for LXC containers that work with Canonical’s Ubuntu Linux OS for Part One of LXD, the Hypervisor for LXC Containers. Let’s continue with Part Two.

LXD, Some Technical Detail

LXD runs on any one of the number of hosts. “And that is in addition to listening on a local socket, which is, by default, what the LXC command will look for, a local LXD,” says Kirkland.

LXD is an externally accessible system that tools can call out to using best practices and the REST API in order to ask LXD to create a system and start it, stop it, back it up, clone it, mod, and migrate it. “These operations are accessible over the local-command line or via the REST API,” says Kirkland.

Goals for LXD

Canonical is developing LXD with sponsorship of LinuxContainers.org and leadership of the Linux containers project. LXD is a sort of sub-project within the LXC project.

“With LXD, we’ve really set out to provide a hypervisor that looks like an ESX, that looks like a KVM, that carves off resources from a physical machine, or a virtual machine,” says Kirkland.

The versatility of LXD together with LXC containers is that they are as compatible inside VMs and for that matter any environment in private, public, or hybrid Cloud computing scenarios as they are on bare metal. “That is something those VMs, KVM, and QEMU were never able to provide with great efficiency,” says Kirkland.

LXD Releases & Adoption

LXD will use an Apache license. Canonical has maximized the coding staff for LXD and is also inviting outside contributions. Expect LXD to be available in line with the Ubuntu 1510 release, which should appear in October. Expect a solid enterprise-worthy release in time for the 1604, the Ubuntu LTS that comes out in April 2016.

Canonical expects adoption by the customers who asked for LXD, which are the telcos and service providers where canonical has pretty measurable success deploying OpenStack.

According to Kirkland, LXD and system containers map perfectly toward infrastructure as a service and OpenStack where telcos, internet service providers, and financial institutions and banks can benefit from a good, clean, well-tested container platform that they can deploy their system workloads into.

“That was the genesis of the LXD project, ensuring that Ubuntu had a good container technology, enhancing LXC to be able to handle, you know, that REST API interface, those external accesses, cluster those together into a suite of tools that enable clusters of machines to then run containers,” explains Kirkland.

Container Benefits Supported by LXD

The key benefits of these containers according to Kirkland include:

• Containers start up in only a couple of seconds (virtual machines may take 10 to 20 seconds, and physical machines may take minutes to start).
• Customers can stuff container workloads onto the hardware in a much more dense fashion.
• Containers can run anywhere, on any architecture, Intel, ARM, Power, IBM, V-Series, as well as inside of virtual machines.
• Processes, the individual workloads inside of that container, are actually running on the underlying platform.

Benchmarks that include data about LXC workload density metrics show that these Linux containers can house 10-, even 20- times the density of containers that VMs can on a single system. “We see that density, but the real source of that density advantage is, of course, that all of the containers on a given system share a single Linux kernel, whereas with VMs, each VM boots its own Linux kernel,” says Kirkland.

Since those LXC workloads and processes are running directly on actual hardware, they are executing on the actual CPU itself, inside of a physical server, in a native fashion and not through a VM escape or emulation. “Since those processes are literally running on the host kernel, on the host processor, from the end-user perspective, they’re seeing incredible improvements over the latency and performance of their processors, inside of those containers,” exclaims Kirkland.

David Geer

David Geer’s work has appeared in ScientificAmerican, The Economist Technology Quarterly, CSO & CSOonline, FierceMarkets, TechTarget, InformationWeek, Computerworld, Byte.com, ITWorld.com, IEEE Computer Society’s Computer magazine, IEEE Distributed Systems Online, Government Security News, Laptop, Smart Computing, Technical Support, The Hosting Standard (Canada), TechWorld.com (UK), SIGnature, Processor, and the Engineering News-Record. David served as a technician at CoreComm in Cleveland, OH prior venturing into writing.

David Geer has 24 posts and counting. See all posts by David Geer