Software-defined storage: Making sense of the data storage technology
A comprehensive collection of articles, videos and more, hand-picked by our editors
By now you may have heard the term "storage hypervisor." You probably don't know exactly what it means, but that isn't your fault. Vendors that use the term to describe their products disagree on the exact meaning, although they mostly agree on why such a technology is useful.
A vendor panel at the Storage Networking World (SNW) show in Santa Clara, Calif., last month set out to define storage hypervisor. The represented vendors sell different types of products, though. The panel included array-based virtualization vendor Hitachi Data Systems Corp., network-based storage virtualization vendor IBM, software SAN virtualization vendor DataCore Software Corp. and virtual machine storage management vendor Virsto Software Corp.
Can all of these vendors' products be storage hypervisors? It's more accurate to say that, taken together, the storage hypervisor products make up an overview of storage virtualization under a new name. And that new name is already giving way to a newer term. "Software-defined storage" was used interchangeably with "storage hypervisor" during the SNW panel.
DataCore Software Corp. CEO George Teixeira said his company was ahead of the current trend when it started back in the 20th century with the premise that software gives storage its value.
"Today we have fancy terms for it like software-defined storage, but we started DataCore in 1996 with a very basic [PowerPoint] slide that said, 'It's the software that matters, stupid,'" Teixiera said. "And we've seen storage from the standpoint of really being a software design."
Teixiera said any talk of a storage hypervisor must focus on software.
"Can you download it and run it? And beyond that, it should allow users to solve a huge economic problem because the hardware is interchangeable underneath," he said. "Storage is no longer mechanical drives. Storage is also located in flash. Your architecture can incorporate all the latest changes, whether it's flash memory or new kinds of storage devices. When you have software defining it, you really don't care.
"Just like with VMware today," said Teixiera, you really don't care whether it's Intel, HP, Dell or IBM servers underneath. Why should you care about the underlying storage?"
Hitachi Data Systems Corp. is known for its storage arrays, but its chief scientist, Claus Mikkelsen, agreed that software is the key to storage virtualization and hypervisors. "The goal is to have software to optimize and automate all these cool things we do," he said.
"It sounds strange for a storage guy to stand up here and say what we're trying to do is use software to reduce the amount of physical capacity, but the storage play has nothing to do with hardware," Mikkelsen said. "When I look at storage arrays, I see them as powerful computers that move data. We have the compute power to do all these things now."
Can storage follow the path of servers?
Ron Riffe, IBM's business line manager for storage software, said the term storage hypervisor refers to a set of value points similar to those that a server hypervisor provides for servers. These include the ability to move resources around to different hardware with a single set of capabilities and to apply intelligent management in the software that works across server platforms.
"The term storage hypervisor is used to describe those value points, the things that customers want to do with their storage," Riffe said. "It's being able to have a common management layer with a common set of capabilities and services regardless of the storage hardware."
These capabilities, he said, should include a common management layer and the ability to shrink, grow and automatically provision storage resources. It also requires analytics to show customers when to move data and resources for performance gains.
Mikkelsen said a storage hypervisor should separate a LUN from physical storage -- a goal that storage vendors are increasingly trying to accomplish when used with virtual machines and server hypervisors.
"To get to the point where a LUN is just a logical container of data, a storage hypervisor or virtualization lets you mask the physical characteristics of a hard drive, a flash drive, thin provisioning, auto-tiering [and] performance management, so people don't have to do that," he said. "So separating the LUNs from physicality is a critical goal, and a storage hypervisor is the main platform to get there."
Virsto Software CEO Mark Davis added, "A storage hypervisor needs to treat the VMDK [virtual machine disk] file and not the LUN as the center of the universe."
VMDK, the disk format used by VMware, is an independent virtual machine that tracks how the virtual machine environment is stored. Or, as Davis puts it, "VMDK is just a big honking file that happens to have an entire copy of Windows and other stuff in it."
Business case for storage hypervisor
IBM's Riffe said server virtualization fixed the problem of "too many servers that weren't using all the capacity were hard to provision, and you had people punching buttons on physical machines to get a workload up and running. You had too much hardware and were spending too much money on server hardware. I think that same idea is driving CIOs who have long since made their decisions on server hypervisors to now ask, 'How can I get the same economic and speed benefits on the storage side?'"
Davis agreed that the economic impetus to adopt storage virtualization is strong. "The way an organization buys hardware is drastically different now," he said. "Huge infrastructures are built with $1,000 and $3,000 servers. That's what is going to happen with storage."
All these concepts sound great, but the fact that storage hypervisor is essentially another term for storage virtualization and that it's morphing into software-defined storage tells us this is a concept that has not yet caught on.