Flash storage and what to do with it has been the hot topic in the world of storage for the past couple of years. Solid-state storage has arisen to meet the needs of desktop and server virtualisation projects as well as to power business critical transactional operations.
We recently surveyed the flash array efforts of the big six storage array makers and found a wide range of products, from existing arrays retro-fitted with flash drives through to ground-up designed flash products.
In fact, it is the startups that have led the way in the flash revolution, innovating in terms of technical development and forcing the big six to include flash options in their product lines. They have largely done this by slotting flash drives into existing arrays – though these are to some extent hamstrung by less than optimal backplane bandwidth and processing power – or by buying ground-up designed flash array products.
But a number of key startups are still at the vanguard of the flash revolution. In general, their key advantage is that they have been designed around the speed and operating characteristics of flash memory, with adequate hardware – not built down to spinning disk performance – and software that is optimised to manage the multiple voltages, garbage collection, wear levelling etc that flash memory requires.
Many flash array startups make use of compression and data deduplication techniques to increase the effective capacity of their arrays. This works well with flash storage as there’s no performance penalty from the randomness of I/O that deduplication introduces.
In this article, we look at how the flash array startups have tackled the challenge of incorporating flash memory into a storage array and how they have positioned themselves in the market against other startups and the big six.
Violin is, according to Gartner, the market leader in all-flash arrays with 19% share, though it has to be said it’s a pretty small market.
Violin offers its 6000 series flash arrays with MLC or SLC options with capacities that range from 6TB up to 32TB in 3U units and claimed IOPS spanning 200,000 to 500,000 (MLC models) and 450,000 to 1,000,000 (SLC). Connectivity is Fibre Channel, iSCSI or Infiniband.
Violin Memory uses its own custom Violin Intelligent Memory Modules (VIMMs) – via a technology agreement with Toshiba – that combine flash memory and controller logic and redundant connections to other VIMMs in the array for high availability.
Violin’s vMOS operating systems and vRAID 4-plus-1 parity system stripe data across VIMMs to provide a five-year endurance cycle.
NVRAM cache is located at the VIMM so no additional logic is required for cache coherency between controllers.
Violin also makes 3000-series arrays designed as direct-attached storage.
SolidFire builds arrays aimed at cloud service providers and is iSCSI block storage. It has automation and multi-tenancy functionality and admins can assign storage volumes with different characteristics to different customers.
With a service provider market in mind, its arrays are designed to accommodate a wide range of workloads, not just those that require high performance. Data deduplication, compression and thin provisioning are built in to help lower the cost per GB for operations outside of Tier 1 or Tier 0.
The firm claims the 2U Accela can achieve 250,000 write IOPS and 200,000 read IOPS. Connectivity is Fibre Channel, Ethernet or InfiniBand, and Whiptail nodes connect to each other through InfiniBand for high availability. The Accela is available in capacities from 1.5 TB to 12 TB.
The Invictas scale from 6 TB to 72 TB, and are made up of two 1U storage routers plus up to six 2U storage nodes. Capacity and performance scale by adding storage nodes.
Whiptail Technologies added an SME version of its all-flash array platform in June this year. The WT-1100 is a 1U form factor device that runs on the same Racerunner operating system as the Accela and Invicta arrays. The WT-1100 will come with 2 TB or 4 TB of MLC flash and will provide up to 100,000 IOPS. Connectivity is iSCSI but Fibre Channel is planned.
Pure Storage with its Purity operating system is on its third-generation all-flash array. The FA 400 uses dual eight-core Intel Sandy Bridge processors to deliver up to 400,000 IOPS and up to 23 TB of raw flash capacity. Pure uses inline deduplication and claims customers can get up to 100 TB of usable capacity from the FA 400.
Pure Storage uses multi-level cell flash in its arrays from Samsung Electronics, one of the startup's investors.
Kaminario K2 uses commodity hardware and its SPEAR (Scale out Performance and Resilient Architecture) OS to build all-flash nodes that use MLC flash. They come in K-Blocks that range from 6TB to 24TB and can be clustered. Each K-Block includes dual processors, two host ports per node and between four and eight Fibre Channel ports or 10Gbps iSCSI ports.
Kaminario's Flash Endurance Booster extends the warranty offered by the drive suppliers, which is typically three to five years, to seven years. Kaminario says it spreads writes around and has a write-buffer to prevent hotspots.
Nimbus E-class hardware scales from 10TB to 500TB while its S-Class arrays scale from 2.5TB to 100TB. Both are multiprotocol arrays (ie, block and file access) and claim up to 800,000 IOPS.
Third generation arrays in the Nimbus family are the Gemini, which pack 48TB and scale into the petabytes with 1,000,000 IOPS per 2U node.
Nimbus Data Systems’ enterprise flash modules (EFMs) are custom-made MLC hardware that work with its Nimbus’ S-Class and E-Class storage arrays that were launched in 2010. The EFMs have dedicated controller software and local NVRAM. Writes are confirmed after completing to this faster RAM rather than waiting for data to be committed to main flash memory.
Connectivity on all Nimbus arrays is Fibre Channel, iSCSI or Infiniband.
Tegile Systems added is mostly noted as being a hybrid flash array startups with some data acceleration smarts. But the company added an all-flash array to its lineup of Zebi hybrid flash multiprotocol storage arrays in August 2012.
The Zebi HA2800 all-flash array has 4.4 TB of capacity and can handle 200,000 IOPS. Tegile offers HDD expansion shelves that can expand the HA2800 to 146TB of hybrid capacity, with customers able to use the flash array as a cache in front of the spinning disk.
Tegile uses flash as cache and a ZFS-based OS tweaked to provide data deduplication; compression; RAID enhancements; and a performance-boosting feature called MASS, or Metadata Accelerated Storage System. MASS sees ingested data dealt with via metadata headers rather than the full copy, and these are kept in cache or SSD tiers.
Skyera came out of stealth with three models. The Skyhawk 12 model has 12TB of raw SSD capacity, the Skyhawk 22 has 22TB while the Skyhawk 44 has 44TB of raw capacity.
The 1U Skyhawk series arrays have 40Gbps and 10Gbps Ethernet network connections.