OpenStack Follows The Datacenter Out To The Edge

It is difficult not to be impressed with the rapid adoption of OpenStack since the open source cloud infrastructure software platform was first released almost 10 years ago, and that adoption has accelerated over the past few years. According to Jonathan Bryce, executive director of the OpenStack Foundation, two years ago OpenStack was running on 6 million server cores in environments around the world across the sites that the OpenStack Foundation could identify and verify. That number has now hit more than 10 million cores through both the expansion of existing clouds and the addition of new customers.

“Those new deployments have brought on additional scales and additional net-new compute cores,” Bryce tells The Next Platform. “It’s also been in existing environments that have continued to scale. Some of the longtime users like eBay and Walmart and Yahoo/Oath/Verizon Media are all organizations that started out with thousands of cores and now have grown to hundreds of thousands or millions of cores.”

He also notes that OpenStack had more than 65,000 commits in 2018, a number that only two other open-source projects – Linux and Chromium – have accomplished, and that the market for OpenStack sits at $6.1 billion. The infrastructure software is used in 75 public cloud datacenter and thousands of private clouds in an ever-widening list of industries and verticals, from retail and financial services to public cloud, manufacturing and energy.

OpenStack is littered throughout datacenters around the world and the numbers are impressive. And those numbers could all be swamped by what’s coming in the rapidly distributed – fast-developing – world of the near future, with the rise of the internet of things (IoT), 5G networks, greater mobility, and more compute and storage capabilities being pushed out closer to where data and applications are being generated and used. With billions of smart, connected devices and systems in the world today and tens of billions more coming, OpenStack supporters can begin to think about the platform’s possibilities outside of the datacenter.

As we’ve talked about here at The Next Platform, the action around infrastructure is increasingly out at the edge, and system OEMs like Dell EMC, Hewlett Packard Enterprise (and its Aruba Networks company) and Cisco Systems – as well as almost every other tech vendor – are building out their capabilities to get to the edge as quickly as possible. The amount of money that will be spent on infrastructure at the edge will be multiple times that spent on central datacenters.

That holds a lot of promise for OpenStack, Bryce says. OpenStack has advantage out there right now, but it’s still early in the development of the edge and more work needs to be done. But it’s a key strategic imperative for the community right now.

“When you look at just the trend of our daily lives, we’re doing more and more data generation and processing in our personal environments all the time, in our houses, in our cars and obviously with our phones, and when I talk to telecom operators, that’s putting a different level of strain on their networks than before,” he says. “The idea of edge is not just about having a smart thermostat on your wall in your home, but it’s how do you handle the billions of those kinds of devices in a way that our networks can survive and handle it, so distributing some amount of processing capability out into the world will be absolutely necessary. Most people are going to be surprised what that kind of distributed edge footprint looks like over the next ten or fifteen years. We’re going to have more edge computing nodes than actually are in our consolidated datacenters. Even among the hyperscale providers, it’s going to be that big.”

What the edge infrastructure and environment will look like is still being developed, though OpenStack will have a role, Bryce says. What that role will be still has yet to come into complete focus.

The edge will “more layers to it than what I think some of the common drawings for the edge are right now, where you have small datacenters scattered around. I think you’re going to have layers of datacenters,” he says, noting China Mobile’s process for building out its environment, expanding from central datacenters to traditional telecommunications spaces with regional and city footprints that are smaller than the datacenters, and then out to the neighborhoods, which are smaller still. “When you get to very, very endpoint of the edge, you may not have OpenStack there. You may … only have some of the smaller pieces that do bare-metal management or you may have some lighter weight pieces that are managed remotely from one datacenter that’s closer to the middle. Openstack has a lead in it but because so much of this is new, we really need to pay attention to how the deployments are going within the community and make sure we’re delivering on what the needs are when they emerge.”

The foundation already has work underway to prepare for the edge. Last year the OpenStack Foundation created a working group for the edge and began embracing GPU accelerators. In addition, the StarlingX Project is developing a high-performance edge cloud software stack based on the Wind River Titanium Cloud R5 offering developed and open-sourced by Intel and Wind River. Airship is a framework of open-source tools for automating cloud provisioning and management.

OpenStack also has a strong presence in the telecom industry with such giants as AT&T, Verizon, Sprint, SK Telecom, Deutsche Telekom and China Mobile, all of which will be key players as the development of 5G and the edge unfolds. OpenStack was there when telecoms first started experimenting with network-functions virtualization (NFV) and as that adoption has evolved

“What I see happening is that, in the first round of NFV, a lot of the telecoms were dipping their toes into the virtualization waters for their networks,” Bryce says. “But now, as 5G has been coming along, that’s gotten to the point where there’s really a large-scale investment that’s happening to rebuild and retool these telecom networks and we’re seeing a lot of activity driving OpenStack into those environments. Whether you’re talking about AT&T or China Mobile or SKT, these companies are all involved in OpenStack in a big way and doing large-scale deployments. That’s resulted in some different requirements that have made it into Neutron and some of the other projects.”

Some of the features that will be key as the distributed computing environment spreads also have made their way into Stein, the 19th version of OpenStack that was released this month. Key among them was bolstering the functionality of containers and the Kubernetes orchestration platform as well as upgrading network support for 5G, NFV and edge use cases. They strengthened Kubernetes support will be key: according to an OpenStack user survey, 61 percent of OpenStack deployments integrate containers and Kubernetes, either on bare-metal servers or in virtual machines (VMs).

In Stein, the Magnum Kubernetes installer has improved launch times from 10 to 12 minutes down to five, while with the OpenStack cloud provider, users can launch a fully integrated Kubernetes clusters with support from the Manila, Cinder and Keystone services, according to the foundation. The Neutron networking service targets container use cases with faster bulk port creation.

Also in Neutron, users will be able to manage network segment types dynamically through a new API extension, which OpenStack says is important for StarlingX and edge use cases. Neutron also now treats bandwidth as a resource for network-intensive applications, where having a minimum amount of bandwidth available is important.

The new capabilities in Stein adds to what OpenStack will be able to do at the edge as the environment evolves.

“When you get out to an edge environment, there may be some portions that are fully containerized and you’re only running Kubernetes – or you may not even be running Kubernetes,” Bryce says. “If you really just want to get thin, you might just be running atop a container runtime layer directly on bare metal without an orchestration element at all. Those are the kinds of things we really haven’t figured out yet as an industry and a community, but I do think that one of the benefits that people are finding with OpenStack is that within a single environment, you can manage physical servers, virtual servers and containers, and integrate with container orchestration and run all of that in a virtual machine or on top of bare metal. That flexibility is the real business need right now.”

Sign up to our Newsletter

Featuring highlights, analysis, and stories from the week directly from us to your inbox with nothing in between.
Subscribe now

2 Comments

  1. While the OpenStack growth to the edge is exciting, at the same time the amount of present and future investments in Openstack by these companies/organizations may logically not allow newer/better technologies [eg. synchroknot] to surface.

    Nice article Jeffrey.

Leave a Reply

Your email address will not be published.


*


This site uses Akismet to reduce spam. Learn how your comment data is processed.