Rackspace is upgrading the servers it uses to deliver its OpenStack cloud offering, hoping to become a reliable, high-performance alternative to Amazon Web Services.
Its new Performance Cloud Servers will offer better value for money than Rackspace’s current standard server offering, the company said.
It’s possible the price for the new, higher performance service might come down further, once Rackspace is able to ditch name brand servers. Rackspace has been an early supporter of the Open Compute Project, kicked off when Facebook revealed the design it uses for cheap servers that run its offering. But for this upgrade, Rackspace is buying servers from HP and Dell.
So why didn’t Rackspace use Open Compute servers to execute this update? Because the Open Compute model has to be adapted for the different use case that service providers require and it’s not quite there yet, Scott Sanchez, director of strategy for Rackspace said. Open Compute was designed around the particular use case of Facebook, he said. But a service provider like Rackspace has to run a cloud that can support the many different use cases of all its customers.
“The difference between Facebook at scale and a hosting provider at scale is you don’t control the apps” as a hosting provider, said Erik Carlin director, Cloud Compute Product Line for Rackspace. As a service provider, “you can tailor it much less because you don’t control the apps.”
Sanchez elaborated. “Things like SSDs and like giving more reliability to a single customer’s server is more important to us than to Facebook,” he said. “If one fails for them, they have 99,000 more servers that do the same thing. But if a customer is running in our cloud and it’s their only server, it matters.”
“So we’re taking the Facebook designs and adopting them into the service provider world,” he said.
The executives didn’t say when they expected to be able to start buying Open Compute certified servers. But they said as soon as the company is ready to start using them, it will start ordering them as it adds capacity.
Whether it’s using name-brand servers or Open Compute servers, Rackspace is hoping to set itself apart from AWS. While AWS has architected its service so that nodes can fail but another will pick up the slack, Rackspace is instead gunning for less failure. Its new servers use Intel Xeon E5 processors and have 120 GB of RAM.
The company has also ditched “spinning disks” and has upgraded to all solid-state disks. And, it’s now offering 40 Gbps networking.
There are a couple of longer term items that public cloud users should think about with this upgrade.
One is that Rackspace said it expects to be able to update its SLA. Currently the company offers two kinds of SLAs, one around the control plane, which lets users provision servers, and the other around the data plane. The control plane SLA won’t change with this upgrade, Carlin said. But the data plane SLA might, he said.
Currently the data plane SLA states that in the event of a host failure, Rackspace will have a backup online in an hour. With the new severs in place, the company might be able to offer an uptime SLA, he said. “We want to get experience before we make commitments,” he said.
Also, Rackspace thinks that some customers will be drawn to its upgraded cloud because they don’t have to architect for failure. “We are delivering more reliability and performance and customers can focus their development activity on those things that are differentiating for them, and not have to work around failure,” he said.
It’s one way to try to differentiate compared to AWS. “Everyone is gunning for AWS right now, and performance is one area where competing public clouds feel they can differentiate against AWS,” said Gary Chen, an analyst at IDC.
Finally, businesses that have been looking for a public cloud option that can do a better job of running database applications might want to take a second look at Rackspace, Carlin said. Database apps can really benefit from the 120 GB RAM and the higher speed networking, he said.
Rackspace is planning to fully transition to the new servers but isn’t forcing users to migrate. Instead, it’s pricing the new service so that it works out cheaper and more performant than the standard offering, nudging customers to go for the better deal.
Read more of Nancy Gohring's "To the Cloud" blog and follow the latest IT news at ITworld. Follow Nancy on Twitter at @ngohring and on Google+. For the latest IT news, analysis and how-tos, follow ITworld on Twitter and Facebook.