StorONE is revitalizing the all-flash array with its new All-Flash Array.next (AFAn) product that combines Intel Optane solid-state drives (SSDs) and Intel quad level cell (QLC) 3D NAND Storage Tiers. The launch is a boon for enterprise on-premises storage arrays that are thwarting off hyperconverged infrastructure (HCI) and the public clouds.

Powered by StorONE’s S1 storage software, AFAn uses Optane SSDs as a performance tier and Intel QLC flash as the capacity store, which the vendor calls “the next generation of enterprise storage.” 

“The easiest way to say this is that this is the first upgrade, if you will, to the normal all-flash arrays really in 10 years,” George Crump, StorONE’s CMO told SDxCentral. “Give Pure Storage some credit. They, for all intensive purposes, started the all-flash array movement. But really, beyond that, you really had no legitimate option, especially in the enterprise until this product.”

“This is the first product that is demonstrably faster,” Crump added. “We're probably 200% faster than their fastest solution, but we're probably 50% of the cost.“

Despite being a relatively new face to the storage market, StorONE's software-defined storage (SDS) product S1 Enterprise Storage Platform supports all-flash or hybrid arrays, virtual storage, secondary storage, and cloud storage, to block, file, and object storage use cases, and Fibre, iSCSI, NFS, SMB, and S3 protocols — the startup has over 50 patents. 

The S1 Enterprise Storage Platform software is the special sauce that powers the AFAn to extract the full performance of Intel Optane SSDs, automatically tiering data to lower-cost Intel QLC 3D NAND SSDs without read performance penalty. 

Most, if not all, storage software companies generally leverage some existing code whether that be internal or publicly available, Crump said. “Our guys took the harder approach and basically wrote everything from scratch, rethinking algorithms with a real keen focus on being very efficient.”

AFAn’s four Optane drives form a performance tier and eliminate any need for cache in memory. The Optane drives also demote older data to the QLC tier in large sequential blocks to prolong SSD endurance. It also supports workloads including VMware, Hyper-V, Oracle, and MS-SQL. 

Up until this announcement the startup had been able to deliver a strong performance off of mid-range servers, making the bill of materials less expensive for systems with StorONE’s software on it, said Crump. AFAn takes StorONE’s efficiency with those same mid-range servers to the max with Optane drives to “deliver 1 million IOPS from 3 Optane drives, while delivering 85-90% of Intel’s stated performance.” 

For reference, all-flash competitors such as Pure Storage, Dell EMC, and NetApp need somewhere between 12 to 14 drives at $2000 a pop to get the same performance that StorONE does with three, Crump said, in explaining the company’s $25,000 or more price advantage over its competitors. 

For comparison, Dell EMC’s mid-range PowerStore and Pure Storage’s FlashArray//X products both utilize  Intel’s Optane technology as a cache, according to Crump who previously spent 13 years as an analyst at Storage Switzerland. The problem with using it as a cache is that there is only one drive in the system, meaning if that one drive fails the data is lost, he explained. 

The safest way to get data off of that cache as fast as possible is to do a read cache, and “that's what you've seen most of them do.” 

By doing that, vendors negate “one of the key benefits of the Optane drive, which is a very high write performance,” he said. “If you can write to Optane but have to wait for the TLC drive to also get the write, you're essentially writing at TLC speeds. So because we treat Optane as storage, not a cache, in the shipping units we have four Optane drives, that’s 3 terabytes of what we call ‘upper tier capacity’ or about 2.2 terabytes usable,” Crump added.

Vast Data’s Universal Storage hardware also uses Intel Optane persistent memory chips and consumer-grade QLC NAND for fast access without tiered storage. However, that use case is entirely different and focused on large unstructured data sets, Crump said.