PHOENIX--Storage Trends Editor Alex Barrett and Editorial Director Mark Schlack caught up with Daniel Warmenhoven, CEO of Network Appliance Inc., at this week's Storage Networking World conference. Here's what the estimable Mr. Warmenhoven had to say on a variety of key issues.
On reducing the cost storage. Warmenhoven says that tiered storage is definitely the future, certainly at NetApp and that it's already begun. He points to the company's lower-cost Nearstore product line, noting that many customers are already using it for primary storage, despite NetApp positioning it as nearline or backup storage.
We asked Warmenhoven to weigh in on the simmering debate about whether multitiered architectures are really too elaborate, given the downward price spiral of all disk technologies. His response was that while NetApp can put different disk technologies in a single frame today, users shouldn't focus on the disk, but the service levels they have to provide.
"Imagine a critical database that you have", says Warmenhoven. "You're going to have multiple copies in multiple places, mirrored and replicated. But they don't all need the same service level."
Commenting on EMC's push to make Clariion's multi-tier, he says "The rhetoric out of the Northeast is solving a problem of their own making" in trying to get around multiple product lines with different underlying technologies, some acquired, some homegrown.
We challenged him on the cost of buying multiple boxes as opposed to versatile boxes. Doesn't the high initial cost of a frame suggest that fewer, more flexible boxes are better? His answer is that the high cost of a frame is "a vendor pricing problem, not a COGS [cost of goods] issue."
Does that mean that users should buy different boxes for different tiers, or boxes that can contain multiple tiers? Warmenhoven suggests that's not really the issue, just a question of packaging.
On reducing the complexity of storage. The real issue for users, according to Warmenhoven, is having a unified software environment, being able to easily conduct backup, snap and other storage operations between tiers, whether on one box or many. That's why NetApp is following the path of having a single underlying architecture for their product line, with multiple interfaces.
Following on that point, Warmenhoven says that NetApp's next products will incorporate the Spinnaker clustering technology on top of NetApp's internals. In other words, it will be the Spinnaker client software on top of the WAFL filesystem and large parts of the NetApp OS.
"From a code perspective, 80% of the software will be [NetApp's] OnTap, 20% will be Spinnaker internals," he says.
On high-performance computing. He disputes claims that this new architecture will take too long for NetApp to develop, forcing the company to lose its position in high-performance computing applications.
"Our largest single sector is high tech," he points out. "About 20% of our business is there, and that includes a lot of semi-conductor firms and other high-performance applications."
He says the Storage Grid architecture, which will ultimately deliver increased performance through clustering will not roll out all at once and will provide increased performance in several stages.
On storage convergence. Last year NetApp began to move seriously toward SAN/NAS convergence with arrays that can have Fibre Channel or iSCSI connections. At least 10% of NetApp's installations are now on either of those protocols, roughly split. NetApp has been a leader in providing native iSCSI target devices.
"We're serious about iSCSI," proclaims Warmenhoven. "When a SAN runs on Ethernet, you've truly got converged storage."
He says the uptake on iSCSI is running surprisingly high. NetApp's first quarter sales give them a run rate of 140% of IDC's predictions for iSCSI adoption this year.