NetApp Takes Build-It, Not Buy-It, Approach To FlashRay Flash Storage Strategy6:02 PM EST Mon. Mar. 04, 2013
|Brian Pawlowski (left) and Ty McConney discuss FlashRay and NetApp's flash storage technology.|
NetApp's new FlashRay flash storage strategy, in which the company plans to unveil flash storage arrays built on a brand new architecture, stemmed from a company decision to build the technology from scratch rather than taking the more common route of acquiring it.
The result was a "Skunk Works"-like project to build a flash storage strategy that was so secret almost no one at NetApp knew of its existence, said Brian Pawlowski, a NetApp vice president who last year stepped down as the company's CTO to develop the strategy.
NetApp last month unveiled FlashRay, previously code-named Project Mars, as a completely new architecture designed from the ground up for scalable all-flash arrays.
FlashRay is a purpose-built flash array architecture designed internally at NetApp as a solution with enterprise features and scalability.
When all-flash storage arrays based on FlashRay begin to be available for beta-testing in mid-2013 and start shipping as planned early next year, they will include world-class file and volume management, a scale-out foundation, complete data management capabilities, and application integration features to allow them to be optimized for specific applications.
Prior to the introduction of FlashRay, NetApp already had its Data Ontap operating system, which has traditionally been at the heart of the company's unified storage systems, and the SANtricity operating system, which NetApp got with its 2011 acquisition of the Engenio line from LSI and is the base on which its EF540 all-flash arrays are built.
However, Pawlowski said, the company decided it needed a whole new architecture on which to build all-flash storage arrays.
Data Ontap is an effective solution for hybrid storage, which adds a flash layer on top of traditional disk storage to increase performance, Pawlowski said. And SANtricity is a great solution for building all-flash arrays for running a single application, he said.
"The decision to do another [architecture] was not a flip of the coin," he said. "The question was, how to design a flash architecture, which faces unique challenges that spinning disks do not."
That, Pawlowski said, pointed NetApp towards developing the technology in-house instead of acquiring one of a myriad of flash storage startups, which is what arch-rival EMC did last year with its acquisition of startup XtremIO.
NEXT: NetApp Cites Gaps In All-flash Storage Startups
Developing high-performance, low-latency flash storage arrays is not that difficult, as shown by the large number of startups in this part of the storage industry, NetApp's Pawlowski said.
The challenge is to get a strategy and a technology that brings together all the different facets that enable flash storage arrays to work in an enterprise environment with other storage technologies such as NetApp's Data ONTAP-based storage systems, something the many companies that NetApp looked at do not have, he said.
There are about 50 startup flash storage array vendors, but there are gaps in their capabilities, Pawlowski said.
"One company might have scale-out capabilities, but no management integration," he said. "One company has dedupe, but not scale-out. These are not things you can add on. We would have to retro-fit these if we acquired another company. The retrofitting would have taken the same time for us to build from scratch."
Another thing missing from all the startups is the kind of enterprise support customers expect from high-end storage systems, Pawlowski said.
"Others are looking at flash technology as a technology problem," he said. "We are looking at it as a solution problem. The support strategy is so critical. Startups pay less attention to what they consider secondary things. We look at these as a whole. The support capabilities have to be in there."
Because of the need for support, Pawlowski brought Ty McConney, vice president of product operations, to his team from his previous position as head of NetApp operations in the Asia-Pacific region where he set up the company's support organization there.
McConney said that enterprises are already testing flash storage arrays from all the startups, and at least one big NetApp customers said it would use arrays from one particular startup just because it believed that NetApp would buy that startup.
"Enterprises want performance and speed, but they don't want risks," he said. "They'll buy from the startups, but not deploy them globally because of the risk involved."
NetApp is building FlashRay to win, Pawlowski said. "We'll win on the merits," he said. "We're not going to say, 'We're NetApp, so buy this.'"
Developing FlashRay in-house allowed NetApp to start fresh on developing technology to provide the full range of enterprise storage services such as deduplication, compression and thin provisioning that conserve storage capacity differently from disk-based arrays because of flash memory wear, Pawlowski said.
NEXT: Will Keeping FlashRay In-house Make NetApp Late?
Scott Robinson, president of XIOSS, a Las Vegas-based solution provider and NetApp partner, said it appears to him that NetApp decided to build its flash storage array technology in-house in order to get tight integration with its Data Ontap offering.
"They probably decided it's easier to build its own rather than bolt the integration on later," Robinson said. "Integration with its own Data Ontap is important for NetApp."
However, Robinson said, while NetApp has a lot of really smart people, much of what it is doing with FlashRay, such as using things like dedupe to manage the flash memory wear leveling issue, is also being done by startups.
"I really like Pure Storage, which has in-line dedupe tied in with flash memory wear leveling that helps bring the cost of flash arrays down to that of tier-one storage arrays," he said. "But Pure Storage doesn't have built-in replication or a lot of the software that NetApp offers."
While the flash storage array business is still in the early stages, the concern for a company like NetApp is that customers are going to be making a lot of flash storage decisions this year, Robinson said.
"I expect NetApp will not be the cheapest guy on the market with the kind of support they plan to offer," he said. "Some customers will wait. But, others will look at alternatives."
Once the decision was made to develop flash storage array technology in-house, NetApp set up the development as Project Mars, a separate group headed by Pawlowski, who actually stepped down as NetApp CTO to lead it.
It was a decision Pawlowski never regretted. "As CTO, I was involved in most acquisitions," he said. "But I'm having a great time with this project."
Project Mars was structured as a startup company within NetApp. "It was not set up as a legal entity," Pawlowski said. "It's an isolated group within NetApp."
Pawlowski's team is located in a building near NetApp's Sunnyvale, Calif., headquarters. "Our building has very restricted access," he said. "It's open just for our team. No one knows the building is owned by NetApp."
PUBLISHED MARCH 4, 2013