With its transition from the one-time powerhouse of the cloud platform space to a data center management provider now complete, Rackspace announced Thursday the availability of a fully managed service option for customers deploying OpenStack cloud platforms, including on their own premises, as well as in any data center, anywhere.
In a development that underscores just how dramatic a change of direction this is for Rackspace historically speaking, colocation leader Equinix has stepped in to provide an option for hosting Rackspace-managed OpenStack platforms through its worldwide locations.
“We are taking the OpenStack product and delivering it in pre-engineered, cabinet-level solutions that not only allow us to provide the OpenStack layer, but (also) an OpEx model. explained Darrin Hanson, vice president and general manager of Rackspace’s OpenStack private cloud operations, in an interview with Datacenter Knowledge. "We will own the assets; we will basically host it back to you in your own facility, and all you’re responsible for as a customer is the space and the power in the data center, and making sure it meets the minimum guidelines.”
Your Place or Theirs
Rackspace’s move is the latest step in a path set in motion in 2014 that has since culminated in managed service offerings for customers of VMware, Amazon, and Microsoft Azure; and just last February, customers with Red Hat OpenStack. A move to provide similar managed services for customers on Google Cloud may not be far behind, especially considering that just Wednesday, Google and Rackspace announced their joint development of Open Compute Project specifications for a 48V open rack using IBM Power9 processors.
Yet the most extraordinary element in this turnaround story is the inclusion of Equinix. As Rackspace told us, Equinix will be its preferred third-party provider for colocation. So where Hanson noted that customers will be responsible for space and power, they’ll have an OpEx option available for those as well.
“Rackspace will take care of all the data center and connectivity management in that solution as well,” explained Ryan Yard, Rackspace’s director of solution engineering, speaking with Datacenter Knowledge. “On a customer site, with this solution, they’d have to manage the power and some of the connectivity. In an Equinix solution, Rackspace would manage that power via Equinix and then manage the connectivity. So we’re extending all of our sites and capabilities into all of the Equinix sites.”
This brings Equinix’ Cloud Exchange squarely into the picture. Just this week, Equinix announced it is supercharging its Cloud Exchange service with the addition of Data Hub — a way to bring customer data closer to the edge of connections with over 500 SaaS providers. Thursday’s announcement means that customers will now be able to effectively hire Rackspace to build cloud platforms that connect Cloud Exchange member services, using open APIs, to applications built on PaaS platforms on top of OpenStack. For example,Cloud Foundry and Red Hat’s OpenShift provide those applications to their own customers as services, according to Rackspace.
“Rackspace chose Equinix facilities to deploy this new service,” said Sean Iraca, Equinix senior director of cloud innovation, in a note to Datacenter Knowledge, “because of the breadth of cloud services available on inside Equinix and on the Equinix Cloud Exchange, as well as the global reach of the platform.”
We asked Iraca about his company’s view of the difficulties Rackspace portrayed that its customers may have had with deploying and managing OpenStack in the past. “OpenStack can be complex, and Equinix provides the best environment, together with services from Rackspace, to run an OpenStack deployment,” Iraca responded.
Would an organization that prefers not to manage its own infrastructure fit the profile of a company with developers willing to build custom apps and services, for a PaaS platform running on OpenStack — one they’d probably have to manage themselves?
“Usually if you are using OpenStack, it means you want to orchestrate infrastructure through the OpenStack APIs in addition to PaaS,” responds Gary Chen, IDC’s research manager for Software-Defined Compute. “Most OpenStack apps are custom developed, so (the customer) would have developers on staff. Most are Web-based or partially/fully cloud-native designs, which run on OpenStack well. They could be using PaaS systems, though the customer would have to install and manage that layer.”
But an app that’s very fast-paced and susceptible to rapid change — for instance, one developed using continuous integration / continuous deployment (CI/CD) — would be something you’d find from an organization that’s focused on the product rather than its own infrastructure, said Chen.
So an organization choosing to devote humanpower and resources to differentiated, customer-facing services, could very well fit the profile Rackspace is looking for: one whose growth strategy does not include devoting resources to undifferentiated infrastructure.
Rackspace will remain the single point-of-contact for customers choosing the Equinix option, said Rackspace. However, according to Equinix’ Iraca, “customers should be perceived as ‘joint’ customers of Rackspace and Equinix.”
“OpenStack has always been the first product that Rackspace delivered as-a-service, outside of our data center,” said Rackspace’s Hanson.
“Up to this point, what that has looked like is a design session where we helped you understand what your OpenStack cloud needs to look like, and then if you decided to deploy it in your own data center, our level of consultation was limited to the bill of materials, the network configuration, and instructions or side-by-side assistance with standing up your environment after you worked with your procurement teams to lay out all the capital, buy or repurpose the gear from your existing inventory, and get it all configured. And if it’s in your data center, then we provided support remotely, and from a monitoring perspective, for just the OpenStack layer.”
There was a lot of hand-holding involved, and plenty of opportunity for Rackspace to demonstrate its “fanatical” devotion to leading customers through the morass. Under the new arrangement, as Yard said, Rackspace will go so far as to assist customers in procuring hardware that meets the agreement’s specifications, for hosting within customers’ own data centers or within Equinix locations.
For organizations seeking to deploy their own services for the first time on cloud platforms, the adoption process for OpenStack has been notoriously difficult. As one user complained on OpenStack’s own forums, “What other software requires the person installing it to do so much work?"
“OpenStack is a collection of projects; (it's) not inherently a product,” explained Rackspace’s Yard, in the collection’s defense. “You have various members of the community whom I think of as ‘curators,’ who take different projects and put them together, but they still allow for the customer to configure that project in any way they see fit. I think that’s where, in my mind, the problem lies.”
For a small, static private cloud deployment, he said, the platform doesn’t really provide too many operational challenges. . . up until the time you want to scale, update, or patch it. “Rackspace’s opinion is that this ‘-as-a-service’ model is a prescription for how to deploy and set up OpenStack at scale.”
Rackspace gave Datacenter Knowledge an early peek at those hardware specifications. There will be two main “flavors” of cabinets, we were told — “HPE” and “Open Compute.”
“HPE-flavor” cabinets will consist of 22 servers, networked using a Cisco top-of-rack switch, Cisco firewall, and F5 Networks load balancers. The “compute” version will feature 2 12-core processors, 256 GB of memory, and 12 960 GB SSD drives. By comparison, the “storage” version will feature 2, 6-core processors and 128 GB of memory, with the slack taken up by 10 4 TB HDD drives and 10 960 GB SSD drives.
This configuration, said Yard, enables support for object storage (via OpenStack Swift), block storage (OpenStack Cinder), as well as a converged array based on Ceph, simultaneously within a single cabinet.
“Open Compute-flavor” cabinets are slightly denser, featuring 24 servers built using Quanta’s F06D chassis — which, the manufacturer says, follows the latest Open Compute 3.0 specifications. It utilizes Quanta’s “hyper-converged infrastructure,” which the company premiered last year in an effort to address the needs of OCP-compliant customers. In that design, compute and storage resources are co-resident. Buildouts, we’re told, will be slightly different than for “HPE-flavor.”
Red Hat — which began certification of its own version of OpenStack when managed by Rackspace — had yet to make a statement about the possibility of certifying OpenStack Everywhere deployments at the time of this filing. A Rackspace spokesperson told Datacenter Knowledge that Red Hat will not be certifying deployments, at least at this stage.