Kenneth Hui, open cloud architect at Rackspace, isn’t a big fan of the term software-defined storage — especially when discussing the cloud. He prefers “programmable storage” to describe capacity that is flexible enough to expand and contract-based traffic workloads and the resources needed.
“It’s gotten to the point where software-defined means anything to anybody,” said Hui, during an interview at the Virtualization Technology Users Group (VTUG) last week at Gillette Stadium in Foxboro, MA. “In OpenStack, we have goals to make storage programmable. It’s programmable in the way it’s consumed. I don’t want to go to the storage team to provision storage. It’s managed by a team of cloud administrators and requests are put in by the end users.”
Hui was part of a two-man speaker team (the other was Cody Bunch, a cloud/VMware expert and author who addressed the audience barefooted) that delivered at keynote at VTUG about cloud principles and bridging the gap between VMware and open-source OpenStack.
“You have to understand, OpenStack is not a virtualization tool,” Hui said. “It’s not a month-long software project. It’s a collection of software projects. OpenStack puts things together to create a cloud platform. It’s a new management layer. It’s one orchestration tool where you spin up enterprise infrastructure from a single pane of glass.”
In the OpenStack cloud world, storage is a part of the overall infrastructure but not a one-size-fits all configuration for every application and traffic workload. OpenStack Swift is object storage for the pure cloud applications that need to scale to petabytes of data. Wikipedia, which uses Swift, is an example of a cloud application that requires object storage.
There also is OpenStack Cinder for persistent or block-based storage for high-performance application, while OpenStack Compute uses ephemeral storage that is like a data store that is created on the fly and then deleted when it is no longer needed.
“In cloud, instances are mostly temporary,” Hui said. “You usually spin up an instance to fit a specific requirement and you adjust the resources to fit the workload. Right now, in the traditional data center the storage guy tunes the storage and then hands it to the compute guys.”
The bigger question is whether traditional, monolithic storage will co-exist with t cloud or become marginalized by it. Hui believes in the former rather than the latter.
“It’s like when the open systems guys said, ‘This is the end of mainframes.’ Where are mainframes today?” Hui asked. “Mainframes generally make more revenue today than cloud. There will always be some workloads that will stay on mainframes, and it’s not trivial to move off legacy systems. It’s never going to be trivial. When I talk to storage administrators about cloud storage versus traditional storage, it’s not an either/or conversation. It’s what is the best use case.”