Mesos 1.0 brings a new container runtime and more third party integrations


The first round number release of Apache Mesos launched into general availability this morning with a slew of new features for data center operators. The arguably most notable addition is a “unified containerizer” that provides the ability to centrally provision Docker, rkt and appc instances.

Removing the need to use each technology’s native runtime provides a couple of major benefits. First, it reduces the number of moving parts that Mesos users need to worry about. And second, the new feature spares the project’s developers from the hassle of adding interoperability with those runtimes, which is much more complicated than merely supporting their respective container formats. As a result, the team should have more time to spend on other activities, like developing features that can broaden the data center operating system’s appeal.

One example of such a capability is the new Nvidia GPU support that is rolling out alongside the unified containerizer. The addition makes Mesos better equipped to run graphically-intensive workloads such as data visualization software, modelling tools and simulations. The chip maker’s processors are also popular among organizations developing artificial intelligence applications, which usually require much higher core density than what general-purpose x86 CPUs from the likes of Intel Corp. provide.

The GPU integration is joined by support for Windows that is aimed at making Mesos attractive for an even wider range of use cases. According to today’s launch announcement, the port is only partially functional in its current form, but full integration is slated to arrive later this year. Mesos co-creator Ben Hindman told InfoWorld that there are also plans to make the platform work better in large multitenant environments that are spread over multiple geographic regions.

Slowly but surely, the framework is emerging as a force to be reckoned with in the software-defined data center. Mesopshere Inc., the startup that Hindman started with his colleagues to monetize their project, has already landed deals with major brands such as Verizon Communications Inc. and Bloomberg L.P. and is showing no signs of slowing.

Image via Wikimedia