The rise of cloud storage, hybrid cloud and software-defined storage have all contributed to a sense that hardware is no longer what is important. In other words, it’s the data that’s precious, while the processing is disposable.
One company that aims to carve out a hardware-centred niche in all that is UK-based SoftIron, which assembles its storage hardware down to sub-assembly level in-house. It supplies Ceph object, file and block storage on its HyperDrive appliances and claims big savings on power, cooling and hard drive life in its heavily optimised products.
But who is going to want all that, in this age of compute and storage as a commodity? We found out during a recent IT Press Tour presentation.
The theory behind SoftIron’s strategy is based on a recognition that the world of IT is becoming ever more dominated by commoditisation. That is, the move towards the cloud as a utility and the software-side emphasis of software-defined means hardware will, paradoxically, be boosted in value. But only if the value of the data dictates it.
That was the view set out by SoftIron CEO Phil Straw, who said: “If data is precious and processing is disposable, infrastructure will go to where the value is greatest. So, edge hardware will have real value.”
That’s SoftIron’s strategic premise – that if the edge is important, customers will want to take extra care of their data there, especially if the edge means locations that lack management, security, power and cooling resources and with conditions that are generally unsuitable.
At the same time, SoftIron is keen to stress the reasons customers can be disenchanted with the cloud.
Strategy vice-president Andrew Moloney suggested the market is settling on a hybrid model. In other words, customers will use the cloud where they can, but at the same time they don’t like egress costs, latency – “it doesn’t make sense to keep data where performance is an issue” – and the fact that promises of multicloud agility have not come to fruition.
Softiron’s response to all this is to build – and it really does build its own hardware, right down to electronic sub-assembly level – Ceph-based open source storage appliances in which all components are thoroughly vetted for security.
SoftIron claims that its products – based around Hyperdrive storage, with routers and switch products also – provide much better performance than those of more commodity-minded rivals because arrays are optimised to achieve synergies between operating system and processors that are simply glossed over in mainstream IT manufacturing processes, while further efficiencies not available to bulk manufacturers reduce wear and tear to hardware.
“There are lots of compromises in the storage design process,” said COO Jason Van der Schyff. “The same 1U appliance could end up as a web server or a database server. Often that might have no effect, but with Ceph the hardware makes a difference, such as when acknowledging reads and writes. With Ceph, it’s not a compute problem, but an I/O problem.”
SoftIron makes a big thing of making “task-specific” hardware, and has optimised its ARM Seattle CPU-based hardware to the Ceph storage software it deploys.
That efficiency of processing led to a claimed 2x saving per dollar per U of rackspace against Dell storage in benchmark testing, as well as electrical and HVAC costs on SoftIron hardware coming in between one-fifth and one-tenth of rivals. Meanwhile, tackling HDD vibration by regulating drives to spin at different speeds to eliminate harmonics has led to drive life doubling in MTBF terms, said Van der Schyff.
As one would expect, SoftIron’s market seems quite narrow. Although it is keen to talk of edge use cases and has targeted the media and broadcast vertical – “a big opportunity” – with its HyperCast video transcoding appliance, customers are mostly in the supercomputing/high-performance computing space.
CEO Straw said deal sizes average 3PB to 5PB and customers made public include Minnesota Supercomputing and the University of Kentucky, although he added that the company had seen “a shift from universities and research to the enterprise”.