Start-up Pure Storage today emerged from development mode to announce an all-solid state drive (SSD) storage array that it said challenges traditional hard drive-based systems on price and blows them away on performance.
Pure Storage claims its FlashArray (FA)-300 series arrays are 10 times faster and 10 times more space efficient than disk-based arrays, reducing the amount of inline data by up to 20 times through deduplication and compression algorithms. The rack-mountable array also uses one-tenth the power of a traditional storage array, the company said.
The FA-300 series use Samsung consumer-grade, multi-level cell (MLC) flash-based SSDs, but the company adds its own clustered array controllers, with an active-active configuration, managed by its own software called Purity. The software offers the ability to thin provision capacity to applications, meaning it can supply additional capacity on a need-only basis instead of the more traditional practice of overprovisioning.
The arrays include the FA-310 and FA-320 models, which respectively offer up to 5.5TB and 11TB of raw capacity natively, but can also double their capacity with expansion trays. Using deduplication, Pure Storage claims the arrays can store the equivalent of between 25TB and 200TB.
The FA-300 series flash array
A single 2U (3.5-in high) FlashArray can achieve a maximum of 300,000 read I/Os per second (IOPS) and up to 180,000 write IOPS. Measured another way, the FA-310 can achieve up to 2GBps sequential read and 1Gbps write rates. The model 320 can achieve up to a 3GBps read rate and a 1.5GBps write rate. A company spokesman said the arrays can sustain the latter I/O rates 24 hours a day for five to seven years.
The company said it can offer SSD capacity at $5 per gigabyte based on an 11TB array, but that pricing is achieved through the use of a 5:1 data reduction ratio; it's not based on raw storage capacity. The pricing also includes data RAID overhead and high availability (HA) overhead. The company refused to release actual pricing for the array, saying it was not "finalized."
Based on the $5-per-gigabyte figure, an 11TB array would retail for $55,000.
The Pure Storage said its FlashArray delivers HA by automatically rebalancing application workloads across remaining controllers in the event of a controller failure, including the completion of any writes still residing in NVRAM.
The company is aiming the all-flash storage array at mainstream corporations dealing with increasing application workloads due to server virtualization, virtual desk top infrastructures (VDI), databases and cloud computing.
Matt Kesner, CIO at the lawfirm of Fenwick & West in San Francisco, said his company has experienced "amazing results" using Pure Storage in its data centers. "The FlashArray has reduced our data between 50% to 90% on a variety of workloads, ranging from VMware virtual machines to Microsoft Exchange and SQL, as well as reduced our physical storage footprint far beyond our expectations," he said in a statement.
Steve Duplessie, senior analyst and founder of market research firm ESG, said Pure Storage has set the storage system bar by offering a "realistic potential for price advantage over regular enterprise disk.
"I love the idea that Pure Storage is aiming beyond simply 'Why not?' as the question users would ask regarding all-flash arrays - pretty darn good in itself. Instead, they are looking to turn the question on its head and make people start to ask, 'Why disk?'"
The FA-300 series offers both 8Gbps Fibre Channel and iSCSI protocol ports for connectivity. The model 310 comes with four ports; the model 320 with 8 ports.
The FA-310 uses a maximum of 650 watts of power and the FA-310 peaks at 1,300 watts. The FlashArray product also comes with fully redundant power and cooling supplies and call-home remote support.
The Pure Storage FlashArray is powered by processors with 12 CPU cores, allowing it to offload deduplication and compression workloads from the I/O stream. The array performs with less than one millisecond average latency with deduplication turned on.
Pure Storage claims that by using data reduction algorithms to remove 70% to 90% of the write I/Os from the data path, the array's write performance also improves because of increased bandwidth. The deduplication algorithm also does not rely on hashing to validate duplicate data. All data on the Pure Storage FlashArray is compared byte-for-byte before being deduplicated because NAND flash excels at reads.
By using read comparisons to deduplicate data, the array avoids the risk of hash collisions, which can occur when two pieces of data have the same hash value. Hash collisions can result in data loss.
Purity's data reduction is carried out throughout the entire array, working down to 512-byte segments, which is what gives it the 10-fold data reduction over approaches that stop at 4kb segments.
While Samsung's SSDs use error correction logic on their controllers to ensure data integrity, Pure Storage's Purity software provides an additional level of validation as all data is also check-summed as it enters the FlashArray. The checksum is validated at every layer of the architecture as data is processed and before being returned as part of a read request.
Lucas Mearian covers storage, disaster recovery and business continuity, financial services infrastructure and health care IT for Computerworld. Follow Lucas on Twitter at @lucasmearian or subscribe to Lucas's RSS feed. His e-mail address is email@example.com.
Read more about storage in Computerworld's Storage Topic Center.
Join the CIO Australia group on LinkedIn. The group is open to CIOs, IT Directors, COOs, CTOs and senior IT managers.