Openstack community releases Queens with support for vGPUs

Everything you need to know about the latest – and 17th – version of open source infrastructure software Openstack, Queens.


The latest – and 17th – version of open source infrastructure software Openstack, named Queens, is now available.

Six months on from the previous release, among the new features in Queens is full support for virtual graphic processing units (vGPUs) in the Nova provisioning component, so if a user is running Nova cloud or has physical servers with GPUs in them, those can now be tracked or provisioned out.

openstack queens hires

Queens also introduces Cyborg, a tool for managing hardware and software accelerator resources including GPU, FPGA, and DPDK/SPDK. It's expected to be useful for telcos running NFV workloads where they will be able to attach and detach accelerators to an instance on the fly.

"The use case for that is thinking more broadly," says Jonathan Bryce, executive director of the OpenStack Foundation. "So, for telcos that have a lot of hardware accelerators in their networks to handle video and mobile data, and that kind of thing."

The new Cinder Multi-Attach feature allows operators to attach the same Cinder volume to multiple VMs – meaning if a node goes down another can have access to the volume. The redundancy supports high availability for mission critical workloads, which the Foundation describes as a "largely unmet challenge" in cloud until now.

"Cinder Multi-Attach lets you create a block storage device and connect it to multiple virtual machines, and so this replicates functionality you used to have to spend gobs of money to get," says Bryce. "Things like storage systems, fibre channel arrays.

"Now you're able to create very highly available clusters using virtualised storage and software-driven storage systems, so this is a big feature for supporting enterprise workloads but doing it at a fraction of the cost."

Queens also sees the first official release of Openstack Helm for managing Openstack lifecycles on top of Kubernetes, and is currently rolled out at American telecom giant AT&T and in pilot at South Korea's SK Telecom.

"I think it's one of the biggest advancements of ease-of-use and operations we've had in years," says Bryce. "We're using containerised Openstack services to manage the lifecycle of the cloud... it's just made a ton of progress in the last six months."

A feature that didn't make it into this release is Fast Forward Upgrade, but this is targeted for the next release, Rocky.

Fast Forward Upgrade will allow users to skip releases in their environments in terms of upgrades, so not just from one release to the next.

"That is another important component of operations and management that didn't make it into Queens, it's something that we hear a request from companies who don't necessarily want to upgrade every six months.... They might be on a yearly upgrade cycle so it helps to address that use case."

Other improvements in Queens include instance repair for bare metal in Ironic, the bare metal provisioning tool.

The community has also brought together a whitepaper on edge computing. During the last Summit, in Sydney, Australia, the Foundation talked about how it wanted to organise work around specific use cases, and edge is one of the biggest for the open source infrastructure.

Edge is where the computing is done on the outer reaches of the network typically on a device. This could mean for connected cars or IoT devices, for example – anything where it makes more sense to compute close to the device and for lower latency. 

See also: What is edge computing?

The whitepaper (PDF) sets out to examine the use cases – "if you look at that continuum of core data centre out to IoT device, it's really focused on the middle of that space" – so where the data lives as soon as it's collected from a car, a thermostat, a VR headset, and so on.

"What kind of capability needs to exist in that space?" Bryce asks. "It's a very new area where I think there's a lot to be defined.

"I think what we're seeing is there are going to be Openstack components, and components from other projects, things like Kubernetes for sure," Bryce says. "The overall takeaway is that there's a huge opportunity for open source in the edge space but we have a lot of work to do to tie it all together."

"Recommended For You"

Openstack Foundation branches out to 'open source infrastructure' Openstack moves one step closer to the edge