This new release delivers high availability for virtual machines, plus storage, security, and networking features enabling increased performance for I/O-intensive workloads
Introduced just six months ago, Mirantis OpenStack for Kubernetes represents the culmination of years of engineering experience building, deploying and operating open source cloud solutions. The result, in short, is the most consumable, operationally docile, and performant OpenStack distribution available.
Cloud operators are using MOS to host virtual machine workloads including mission-critical enterprise applications, databases, and telco virtualized network functions. MOS can also host Mirantis Container Cloud, enabling deployment of consistent Mirantis Kubernetes Engine clusters on MOS, in just a few clicks.
Introducing MOS 21.2
Now, Mirantis OpenStack for K8s (MOS) 21.2 is available, and some users are already consuming the update via the continuous delivery mechanism. The latest release continues our past few months of rapid innovation around Network Functions Virtualization, and enhances the Tungsten Fabric backend for OpenStack Neutron with network acceleration capabilities, including single root input/output virtualization (SR-IOV — a standard method for virtualizing and sharing compatible PCIe network cards directly with guest OS) and improved support for Data Plane Development Kit (DPDK — a programming toolkit for offloading TCP packet processing from kernel to userspace processes).
In combination with OpenStack’s CPU pinning, huge pages, and NUMA-awareness capabilities, these turn MOS into a strong foundation for any Network Function Virtualization platform. Tungsten Fabric, meanwhile, becomes now an SDN option not only for clouds based on OpenStack Ussuri, but also for greenfield deployments with OpenStack Victoria.
Instance HA aaS
Release 21.2 now introduces a long-awaited feature: Instance High Availability as a Service (Instance HA aaS). The first version is capable of recreating a VM on the same host in case of a system failure, and now we are working on enabling more complex use cases, like automatic evacuation of instances from a failed compute node.
High Performance Root / Ephemeral Storage with Encryption
IO-intensive workloads can now achieve improved performance by hosting VM root and ephemeral storage directly on the host, rather than on Ceph distributed storage. This is especially relevant for the applications that bypass Linux page cache when writing or reading data.
To enable use of local storage, a cloud operator allocates a Logical Volume Manager (LVM) group when provisioning a compute node and gives it away to OpenStack Nova completely so the latter can dynamically allocate volumes on users’ demand. Users relying on LVM-backed virtual disks can also enable on the fly encryption of data using the key storage in the secrets manager, OpenStack Barbican.
By combining high performance (LVM) and highly-resilient (Ceph) storage, a single cloud can satisfy a large variety of high-performance computing use cases.
Deployment and Updates
behind a Proxy
MOS is now capable of getting deployed and receiving updates, even in enterprise environments where access to the Internet is restricted. In order to fetch artifacts, a MOS cluster will connect to Mirantis CDN via a secure corporate HTTP proxy server, optionally protected with authentication.
Alternatively, the Mirantis Container Cloud (MCC) management cluster can also act as a proxy, in addition serving as a cache for the binaries.
Introduced together with the Release 21.2, TryMOS is a new and simple way to try out Mirantis OpenStack for Kubernetes. It is a special flavor of MOS, which is distributed as an all-in-one Amazon Machine Image (AMI). While the miniaturized version (obviously) does not provide remotely the same performance and feature-set as a full-scale MOS cluster deployed on bare metal, TryMOS still lets you explore OpenStack and the lifecycle management mechanisms Mirantis has built around it.
Just click to Try MOS today!