SoftIron's strategy to bring Ceph storage to the masses: 'A really, really sh*tty computer'...

... but a great storage appliance, which is the point, says CTO Phil Straw

Storage startup SoftIron has launched three products designed to make it easier to deploy Ceph – a free software storage platform that supports block, object and file storage.

New toys include an unusual 1U server based on ARM64 architecture; a gateway appliance that gets Ceph to play nicely with legacy enterprise environments; and HyperDrive Storage Manager software that complements Ceph and aims to simplify the work of sysadmins.

SoftIron describes Ceph as the "Linux of storage". Designed to run on clusters of commodity servers, it offers a software-defined storage backend at virtually no cost, complete with replication and snapshot functionality.

However, Ceph remains a relatively young open software project – the first major stable release arrived in 2012 – and the community distribution is notoriously hard to deploy and configure. Vendors with their own distributions of Ceph include Red Hat and SUSE – both charge a subscription fee, and neither bundles integrated hardware.

"If you want to run Ceph you have to be a bit of a propellerhead," SoftIron CTO Phil Straw told The Reg. "I'm not un-technical, I do technology for a living, and I remember learning Ceph was very intimidating."

SoftIron has been working on Ceph since being founded in 2012, making its own additions to the core platform. The company made headlines as one of the first vendors to produce servers based on ARM64 architecture. It operates its development and manufacturing in Silicon Valley, but is registered in the UK, with headquarters in Mayfair.

SoftIron's latest server is the HyperDrive Performance Multi-Processor (HD32112), which manages to squeeze two storage nodes, with one or two CPUs each, into a single rack unit. This being Arm, the compute component is somewhat underpowered, and the hardware design uses several engineering tricks to improve drive performance.

For starters, it reverses the traditional storage server layout by moving CPUs to the front of the PCB and storage drives to the back. This means cool air from the fans blows over the drives first, and then the CPUs – which wouldn't make any sense in a compute server.

Among other reasons, the machine uses low-power CPUs – SoftIron claims the system consumes less than 200W under full load – to keep the internal temperature of the server cooler which in turn helps make the drives last longer. Straw claimed that, on average, SoftIron servers run 25°C cooler than a comparable system powered by Xeons.

At the same time, lower temperature enabled the designers to decrease the number of fans, reducing vibration – and vibration is the enemy of HDDs. "Once we worked out that what you actually need to do to optimize for input and output, as opposed to compute – we started with this HyperDrive product," Straw said.

"What we ended up with is an optimized multi-processor design for storage, but a really, really shitty computer. We basically have separate islands of processors, two or four, and because of those processors we have massive amounts of IO that we can directly connect [to drives].

"It's a totally shitty computer, but what we are trying to do here is storage, and not compute, so when you look at the IO, when you look at the buffering, when you look at the data paths, there's amazing performance – we can approach something like a quarter of a petabyte, at 200Gbps wireline throughput.

"We've done something that's very different, very contrarian. You will never find Dell or Supermicro going anywhere near this architecture."

To look after its servers, the HyperDrive Storage Manager software enhances Ceph with features valued by enterprise users – like an actual graphical user interface. It can update, monitor, maintain and upgrade HyperDrive nodes, supporting all the mainstream storage protocols: file shares (SMB/CIFS and NFS), object stores (S3 and Swift), and block devices (iSCSI and RADOS). The company plans to open-source the code in the foreseeable future.

The final piece of the puzzle is the Storage Router – essentially a network gateway that links together HyperDrive nodes and the rest of the customer's infrastructure, and translates various storage protocols.

"Ceph is an amazing technology. One of its properties is that is scales like nothing else. It has these algorithms – RADOS and CRUSH maps – the core technologies inside Ceph that allow it to scale. That's fantastic for bulk storage, but how do you consume it, how do you use it?" Straw asked. "There's managing it – what Storage Manager is for – but how do you actually access it, how do you connect to a Windows machine, how do you connect to an Apple machine, how do you connect to a server that does legacy technologies?

"What we created is a storage router, which allows us to encapsulate all of the legacy, all of the enterprise capabilities that you need to have in order to exist in the real world." ®

Sponsored: What next after Netezza?




Biting the hand that feeds IT © 1998–2019