OpenStack is getting bigger than ever. It now powers more than 75 public cloud data centers and thousands of private clouds at a scale of more than 10 million compute cores. But it's always been hard to upgrade from one version of OpenStack to another, and it's been hard to deploy on bare metals. With OpenStack 18, Rocky, both problems are much easier to deal with now.
The open-source OpenStack cloud, like its ancestors, has always run well on diverse hardware architectures -- bare metal, virtual machines (VMs), graphics processing units (GPUs), and containers. Bare metal was always a bit tricky. OpenStack Ironic, its bare metal provisioning module, is bringing more sophisticated management and automation capabilities to bare metal infrastructure. Nova, which provisions compute instances, now supports creating both virtual machines (VM)s and bare metal servers. This means it also supports multi tenancy, so users can manage physical infrastructure in the same way they manage VMs.
Other new Ironic features include:
- User-managed BIOS settings: BIOS (basic input output system) performs hardware initialization and has many configuration options that support a variety of use cases when customized. Options can help users gain performance, configure power management options, or enable technologies like single root input/output virtualization (SR-IOV) or Data Plane Development Kit (DPDK). Ironic also enables users to manage BIOS settings, supporting use cases like Network Functions Virtualization (NFV) and giving users more flexibility.
- Conductor groups: In Ironic, the "conductor" is what uses drivers to execute operations on the hardware. Ironic has introduced the "conductor_group" property, which can be used to restrict what nodes a particular conductor (or conductors) have control over. This allows users to isolate nodes based on physical location, reducing network hops for increased security and performance.
- RAM Disk deployment interface: A new interface in Ironic for diskless deployments. This is seen in large-scale and high performance computing (HPC) use cases when operators desire fully ephemeral instances for rapidly standing up a large-scale environment.
Julia Kreger, Red Hat principal software engineer and OpenStack Ironic project team lead, said in a statement, "OpenStack Ironic provides bare metal cloud services, bringing the automation and speed of provisioning normally associated with virtual machines to physical servers. This powerful foundation lets you run VMs and containers in one infrastructure platform, and that's what operators are looking for."
This isn't just theory. It works. And it heading into production.
James Penick, Oath's IaaS architect (Oath is AOL and Yahoo's parent company), said Oath is already using OpenStack to manage "hundreds of thousands of bare metal compute resources in our data centers." He added, "We have made significant changes to our supply chain process using OpenStack, fulfilling common bare metal quota requests within minutes."
That's good, but it's not good enough.
"We're looking forward to deploying the Rocky release to take advantage of its numerous enhancements such as BIOS management, which will further streamline how we maintain, manage and deploy our infrastructure," Penick said.
Also: How to install OpenStack on Ubuntu Server with Devstack TechRepublic
That's great, but many OpenStack users are already saying, "Maybe I'll install this in 2021."
Upgrading OpenStack isn't easy. But OpenStack Rocky's Fast Forward Upgrade (FFU) feature is ready for prime time, and it's all set to help users overcome upgrade hurdles and get on newer releases of OpenStack faster. Now, FFU lets a OpenStack on OpenStack (TripleO) user on Release "N", and they can quickly speed through intermediary releases to get on Release "N+3" (the current iteration of FFU being the Newton release to Queens). You can't jump all the way to Rocky, but you can a lot closer to it more quickly than you ever could before.
Other new features are:
- Cyborg provides lifecycle management for accelerators like GPUs, FPGA, DPDK, and SSDs. In Rocky, Cyborg introduces a new REST API for FPGAs. These floating point chips are used machine learning, image recognition, and other HPC use cases. This enables users to dynamically change the functions loaded on an FPGA device.
- Qinling is introduced in Rocky. Qinling ("CHEEN - LEENG"), a function-as-a-service (FaaS) project. This delivers serverless capabilities on top of OpenStack clouds. It also enables developers to run functions on OpenStack clouds without managing servers, VMs or containers -- while still connecting to other OpenStack services like Keystone.
- Masakari, which supports high availability by providing automatic recovery from failures, expands its monitoring capabilities to include internal failures in an instance, such as a hung OS, data corruption, or a scheduling failure.
- Octavia, the load balancing project, adds support for UDP (user datagram protocol). This brings load balancing to edge and IoT use cases.
- Magnum, a project that makes container orchestration engines and their resources first-class resources in OpenStack, has become a Certified Kubernetes installer. This makes it easier to deploy Kubernetes on OpenStack.
Want to check the new OpenStack out? You can download Rocky today.