The OpenStack community issued its Queens software release today. The 17th version of the open source cloud software includes some hefty updates such as a software-defined storage functionality, GPU compatibility, and tracking of container workloads.
The OpenStack software is evolving fast in response to user needs. “Cloud started out as a fast way to get a virtual machine,” said Jonathan Bryce, executive director of the OpenStack Foundation. “Now, people expect OpenStack clouds to solve a lot more problems. The scope of cloud continues to move out and gobble up every possible use case.”
One of the new features in the Queens release pertains to the Cinder storage sub-project within OpenStack. “The Cinder multi-attach feature allows you to take a single block storage device and attach to multiple VMs at the same time,” said Bryce.
According to the OpenStack Foundation, the new software-defined storage capability is one of the most-requested features in cloud environments. The feature provides storage redundancy. If one node goes down, another takes over and has access to the volume.
“Usually, you have multiple front-end servers connected to a high-end server on the back-end,” said Bryce. “If one front-end server goes down, it has access to the high-end server. With Cinder you get the ability with virtual storage to deliver that same level of high availability not dependent on expensive fibre channel storage arrays.”
He said the entry level for a lot of the traditional high-end storage arrays for storage area networks (SANs) begins at $100,000 and can run into millions of dollars.
Storage appliances have always run software, so there are some people who claim software-defined storage (SDS) isn’t anything new. But Bryce said the relatively new term “software-defined storage” refers to the decoupling of storage software from the underlying hardware. “It gives a lot more flexibility versus what we have seen in the last 30 years in storage where vendors have bundled that all together,” he said.
The Queens release also includes support for virtual graphic processing units (vGPUs). Within the Nova sub-project of OpenStack, users can boot VMs that have vGPUs, an important capacity for graphics-intensive workloads.
Mark Collier, chief operating officer with the OpenStack Foundation, said GPUs gained prominence for video game processing, but their algorithms for 3D rendering turned out to be good for machine learning and artificial intelligence (AI), as well. “These chips ended up being widely adopted in data centers,” said Collier.
“It’s great to see this in OpenStack because a lot of organizations are trying to build out these machine learning platforms where they have a fleet of servers with GPUs in them,” added Bryce. He said some early adopters such as eBay and Commonwealth Bank have been using GPUs together with OpenStack. “Before, you could manually configure any hardware in a server and let a VM take advantage, but it required quite a bit of extra work,” said Bryce. “But now it’s native. Queens takes something we’ve seen users doing and makes it available to a much broader audience.”
The Queens release also include a new container networking feature Kuryr within the OpenStack Neutron sub-project. Kuryr tracks the workloads of containers. “If you deploy a workload in Kubernetes you may have no idea what server it’s running on,” said Bryce. “Kuryr watches all that activity and as actions happen, it configures the underlying Neutron network to match whatever is happening.”
The OpenStack Foundation is gearing up for its next summit — this one in Vancouver in May. Some topics such as edge computing and network functions virtualization (NFV) will have their own tracks and get more attention than they have in previous summits.
OpenStack, itself, seems to have recovered its reputation that waivered for a bit the last couple of years when some users complained the software was too complex. “One of the things we needed to work on was better defining the purpose of OpenStack,” said Bryce. “When you start to bring in users and use cases, it makes it very real, and it provided a re-focusing for the community last year. And it clarifies for the market overall what OpenStack is trying to solve for.”
The software now powers 60 public cloud data centers and thousands of private clouds at a scale of more than six million physical cores.