After several years of effort, The Apache Software Foundation (ASF) announced the availability of version 1.0 of the Mesos cluster manager, widely used to manage IT infrastructure for container applications.
Benjamin Hindman, Apache Mesos co-creator and current chief architect and co-founder at Mesosphere, said Mesos 1.0 will give IT organizations access to a stable application programming interface (API) that they can count on being in place for the next several years. In addition, the Mesos API now supports HTTP, so applications don’t have to be changed to support the technology, he noted.
Other new elements of Mesos 1.0 include an engine that supports multiple types of containers image and more granular authentication facilities that make it possible for IT organizations to control who gets access to specific microservices invoked via Mesos.
Finally, the manager is compatible with the Container Network Interface (CNI) and includes support for graphics processing units (GPUs). With Mesos 1.0, companies no longer must rely on Docker daemons that Hindman said often are too buggy or simply don’t scale efficiently in a production environment.
Going forward, the open-source team responsible for driving development of Mesos will focus on improving support for multitenancy and improving the algorithms used to automatically allocate IT infrastructure resources to container applications, he said.
Mesosphere has created a data center operating system (DCOS) on top of Mesos that competes with technologies such as Kubernetes or Docker Swarm. However, Mesos itself is a lower-level cluster manager. Many IT organizations may opt to deploy Kubenetes, for example, on top to better-manage underlying IT infrastructure resources.
Thus far, developers have driven most of the decisions in container orchestration frameworks, Hindman said. But as IT operations teams become more familiar with the challenges associated with running container applications in a production environment, Mesosphere is betting IT organizations will come to appreciate the management functions enabled by both Mesos and DCOS.
Ultimately, Hindman said, developers and IT operations teams will need to strike a balance between the need for developer flexibility and IT operations team requirements when it comes to consuming IT infrastructure in a shared production environment. At present, most IT organizations don’t fully appreciate all the Day Two challenges associated with managing container applications, he said.
Of course, developers and IT operations teams could simply ignore each other. But the better part of valor would be for them to collaborate on a framework that makes it possible for both to achieve what they need in a way that requires the fewest number of interactions possible. After all, neither developer or IT operations staff have a lot interest in spending a lot of time together. Rather, they simply want a set of well-defined rules through which both parties know what to expect from each other without having to go to yet another meeting to get it.