provides the basic software infrastructure that allows corporations to build their own clouds, in much the same way that Microsoft provides the e-mail infrastructure for thousands of organizations through its Exchange server software.
This vision of standardization around a single company’s technology goes by many labels—“federation” is one that Feinberg uses—but the point is to speed up the transition to a world where most computing and storage resources are organized in clouds. (Meaning, at a minimum, that the resources scale up and down with demand and that users pay only for what they need.) The more companies and cloud service providers are running Atmos, and the more vendors of enterprise software tailor their products to work with Atmos, the more incentives there will be to buy into the Atmos vision, Feinberg argues.
“We want to enable an ecosystem,” Feinberg told me during a recent visit to Xconomy. “You will hear us message this again and again. Others will dredge up competition, but we are just trying to be an enzyme, a catalyst to make this whole economy go together.”
Feinberg says he believes EMC has a “second-mover advantage” with Atmos, because it’s been able to watch other companies cope with the new demands created by Web-based delivery of information and applications, learn from their solutions, and design a storage system from scratch that’s tailored for huge amounts of data distributed across multiple storage sites and accessed via the Web.
“If you’re writing an application from a clean sheet today, do you write it for Windows or Linux?” asks Feinberg. “Of course not. You write it for a browser-based experience, whether that means it winds up running on a real browser, or an iPhone, or whatever. It’s an HTTP experience,” referring to the hypertext transfer protocol, the basic data-handling mechanism for the Web. “But truth be told, the existing architectures that the hardware industry provides have not been sufficient to meet that need. If you look at any of the Web 2.0 companies—Google, Facebook—they have all had to invent their own file systems to scale up. That is where Atmos and cloud storage shines, frankly.”
There are a few big differences between traditional storage systems and those optimized for the cloud, according to Feinberg. First, the systems must be able to store petabytes of data in aggregate (that’s millions of gigabytes) rather than terabytes (thousands of gigabytes), and they must be able to keep track of this data even if it’s scattered across dozens of separate sites. Second, they need to be flexible and resilient to change—Feinberg calls this “loose coupling.” For example,
