Modular data center expert IO is getting into the cloud provider business, launching a new service called IO.Cloud that’s built using Open Compute server designs and runs the OpenStack cloud computing operating system.
That’s a lot of open source, but the company seems to think it’s necessary. According to the IO.Cloud website: “IO.Cloud is built on Open Compute because it provides our engineers with the flexibility to configure and optimize the hardware specifically for scale cloud deployments … IO.Cloud uses OpenStack Cloud components that are interoperable and designed to support standardized hardware implementations.”
IO is pitching IO.Cloud as an enterprise cloud offering, and if it plans to legitimately compete against larger cloud providers for those workloads, the company and its cloud can use any advantages they can get. IO.Cloud is available in hosted and on-premises versions, and the Open Compute hardware almost certainly will let IO operate its public cloud infrastructure more efficiently, as well as letting private cloud customers run more efficiently.
Each rack of gear consists of 26 compute nodes and 4 storage nodes, built using the Open Compute Winterfell and Knox designs, and contains 960 terabytes of storage and 3.75 terabytes of RAM. An an entire 18-rack IO data center module can house nearly 17 petabytes of storage and 67.5 terabytes of RAM.
In theory, the hardware and the OpenStack software both allow for flexibility in adapting the systems to changing application requirements, too.
At a higher level, IO.Cloud is designed kind of like Amazon Web Services, albeit at a smaller scale — it’s running in two data centers in the United States, each one representing a site within a geographic region, and each module at each site representing a separate availability zone. The service provides infrastructure as a service and what IO calls data center as a service capabilities. That’s a cloudier version of managed hosting, and data center as a service customers can connect to IO.Cloud public resources via a Layer 2 network connection.
IO is preparing for its initial public offering, and was not able to speak about IO.Cloud because of a company-wide moratorium on press interviews. The looming IPO makes the timing of the new service a little strange, actually — it’s investing presumably a not-small amount of money to get into a business in which it has little experience. The company has done colocation before, but has not offered public cloud resources.
Still, IO.Cloud does show the company is willing to take some chances with what it thinks are big differentiators in its modular data centers and operating system software. Building it on both open source hardware and cloud software is a good way to stand out in an enterprise cloud field where distinguishing among the various offerings can be difficult. And IO does have some idea what large enterprises want: In 2012, it signed a partnership deal with fellow Open Compute supporter Goldman Sachs to supply the bank’s future data center capacity.
Update: This article was updated on Feb. 6, 2014, to remove a reference to a platform-as-a-service offering and to correct the number of IO.Cloud data centers to two from seven. An IO spokesperson said the PaaS offering is a “roadmap service” and has been removed from the product website until it is available. The spokesperson also noted that IO plans to run IO.Cloud in all its global data centers, but that it presently only runs in the company’s Phoenix and New Jersey locations.