Solid state storage pioneer Whiptail Technologies is trying to move ahead of the growing all-flash array pack by upgrading its single-chassis system and adding a modular scale-out architecture. This week Whiptail will launch its Accela and Invicta high-performance storage systems.
Accela replaces the XLR8r flash array that Whiptail has shipped since 2009. CEO Dan Crain said Whiptail has nearly 200 XLR8r units in production. Its systems are targeted at customers with highest read and write performance needs.
“A lot of people are talking about this [flash storage],” he said. “We’ve actually built it and we’re shipping it now. We’re out of the hype zone and focused on substance. We don’t talk much but we deliver. When we talk, it’s about stuff in the field now, not what will be there five years into the future.”
Whiptail claims the 2U Accela can reach 250,000 write IOPS and 200,000 read IOPS with 1.9 GBps of bandwidth and 100 microseconds of latency. The devices connect to servers through Fibre Channel, Ethernet or InfiniBand, and connect to each other through InfiniBand for high availability. The Accela is available in capacities of 1.5 TB, 3 TB 6 TB and 12 TB.
Invicta scales from 6 TB to 72 TB, and consists of two 1U storage routers and up to six 2U storage nodes. The routers provide redundant paths and manage host connectivity and storage pools. The routers also map, mirror and stripe LUNs across nodes. The storage nodes manage data reads and writes requested by the storage router, and manage RAID protection and hot sparing.
Capacity and performance scale by adding storage nodes. A two-node system would have up to 300,000 read IOPS, 250,000 write IOPS, 2.7 GBps read bandwidth and 2.5 GBps write bandwidth. Those numbers rise to 650,000 read IOPS, 550,000 write IOPS, 7 GBps read bandwidth and 5 GBps write bandwidth with six nodes.
Customers can upgrade from Accela by adding it to Invicta as another storage node. Pricing starts at $50,000 for an entry-level Accela.
Although the number of all-flash array vendors has shot up over the past year, Crain said Violin Memory is Whiptail’s major competitor in the field. “We don’t see anybody else,” he said.
Crain said Whiptail has written hundreds of thousands of lines of code to manage wear leveling that makes MLC reliable for the enterprise. Whiptail doesn’t use any of the higher endurance single-level cell (SLC) flash. “We understand the way NAND works and we’ve refined it over the years,” he said. “One of the biggest issues of MLC is around wear, but that’s not a large problem if the system is designed correctly. Most commercial systems don’t stay in production past three to five years, and we can last twice as long as that.”
With its concentration on performance, Whiptail has yet to add management features such as thin provisioning and data reduction. Whiptail CMO Max Riggsbee said those items are on the vendor’s roadmap but its goal is to eliminate any performance impact.
“Our number one goal is to do no harm to performance,” Riggsbee said. “We’re looking at a way to do thin provisioning, deduplication and compression so that customers still get maximum performance whey they want to implement those features.”
The more immediate roadmap includes asynchronous replication, snapshots with merge capability and a VMware vCenter plug-in. Those features are due by the end of June.
James Bagley, a senior analyst with the analyst firm Storage Strategies Now, said Whiptail’s use of InfiniBand as a backbone, the way it protects the write buffer and Invicta’s stretch mirroring across sites stand out among flash arrays.
“Whiptail is aimed at customers with requirements for a lot of IOPS, low latency and read and write speeds that are about the same – that’s where the write buffering comes in,” he said. “Even though they write to flash, the read/write speed is almost identical. VDI [virtual desktop infrastructure] has been a sweet spot for them, along with transaction processing, trading and Web-scale applications.”