Talk of EMC possibly acquiring flash array startup XtremIO last week placed a spotlight on the market for storage area network (SAN) systems engineered around solid-state drives (SSDs). Under that spotlight lies a rapidly growing and competitive group of startups looking to become trailblazers in the journey to flash SAN arrays.
The EMC story served as a coming-out party for XtremIO, which has yet to even formally announce its arrays. There are a bunch of startups that have formally unveiled all-flash arrays, though. Not all of them are shipping yet either, but they are already elbowing for position. Nimbus Data launched its all-SSD arrays in April 2010 and is the graybeard of the group.
There are so many all-flash SAN systems now, they already have sub-groups. There are high performance systems such as Violin Memory, Kaminario, Texas Memory Systems and Whiptail, and those looking to replace mainstream storage arrays such as Nimbus Data, Pure Storage and GreenBytes. SolidFire is targeted at cloud storage providers.
Others are hybrids designed around flash for specific workloads but also have spinning drives for capacity tiers. Tintri, NexGen and Nutanix use flash and hard drives in storage systems built for virtualized environments, and Nimble Storage has iSCSI hybrid systems that handle primary and secondary storage in one box.
There will likely be more flash SANs to come, before the inevitable consolidation sets in.
“I think we’re just scratching the surface on this market,” said Greg Schulz, senior analyst of the StorageIO Group. “You’re going to see shakeouts, and there will be vendors left behind. Some will survive and extend their capabilities. Some vendors are already adding support for high-capacity drives, and features such as auto-tiering and replication to make themselves more appealing as general purpose storage than a niche market.”
The EMC-XtremIO rumors also raised speculation about whether other mainstream storage vendors would look to buy all-SSD startups instead of building their own flash arrays or fully populating arrays built for hard drives with SSDs.
Pure Storage CEO Scott Dietzen called the attention around flash “great reinforcement to one of the founding ideas at Pure -- you can’t retrofit an existing storage array for flash. Building it from the ground up is crucial. The software you have to craft is on the order of VMware’s virtual stack or a relational database.
“We offer a product that is price-competitive with a disk array. We get it down to about $5 to $10 a gigabyte. That was extremely hard to do -- people challenged us whether we can do deduplication and compression fast enough for flash,” He said.
Eight months after announcing its FlashArray product, Pure has yet to go GA with the product. Dietzen said the startup is winding down its early access program and has customers using FlashArrays in production and will identify some of its users in a month or so.
Tintri CEO Kieran Harty, whose company has announced customers, said the challenge for Tintri and the other startups is doing more than building a system around flash.
“We think flash is only part of the equation,” he said. “Our real differentiation is that we sell storage for a virtualized environment. There are only a few vendors out there that can do a good job of being a flash file system, let alone specifically for virtual environments that allows you to do things like replication, snapshots with granularity for virtual machines.”
Schulz agrees that it will take more than flash to make it.
“What makes a system enterprise class? You need a level of redundancy, snapshots, replication, security, authentication -- all things that traditional enterprise storage would have,” he said.
All-flash storage still nascent market
XtremIO executives may disagree, but its rivals say talk of a feeding frenzy with large storage vendors gobbling up startups is a bit premature.
“The notion of a flash feeding frenzy is a little simplistic,” Tintri’s Harty said. “Most players have not come to market yet, so it’s strange that we’re talking about consolidation already. I think consolidation is unlikely for a number of reasons. There’s a large amount of venture capital invested in flash storage companies, and the investors will surely wait for a larger payout. Even $400 million to $500 million is not a home run for a large VC. It’s still early in the rollout of the [enterprise] flash market.”
Dietzen said if startups are counting on getting acquired, a good number will fail. “There aren’t a great deal of players in enterprise storage,” he said. There are about six enterprise storage companies out there, so not all of these startups will make it. We think one or two of us can remain independent. Others that have success will have to do M&A. But it’s hard to have a successful acquisition of an early technology that’s not bearing revenue and proven out in the marketplace. The salesforce of a big company doesn’t know what to do with an unproven product.”