Shared storage costs a lot. It’s probably the biggest single-ticket item in the data centre.
On SearchStorage.co.UK we recently examined the supposed sub-$10,000/euro 6,000 SAN market, and the conclusion was that there really wasn’t one; we determined that the few products available at this price point lacked dual controllers or support and maintenance, and you’d actually need to spend more like $20,000/euro 12,000 to get a decent entry-level SAN.
That’s entry level, remember. Realistically, to deploy a fully redundant SAN with a reasonable amount of options plus maintenance you’re looking at around euro 100,000 for, say, 10 TB of shared storage from the likes of EMC, HP or IBM.
But does enterprise-class shared storage have to be that expensive? Well, right now there isn’t a wealth of affordable shared-storage options. The overwhelming majority of storage products are those produced by the major vendors. They consist of controller software and a bunch of disk enclosures. There’s nothing very special at all about the latter, but the former is proprietary software and is what costs the customer so much.
We’ve seen (and written about in a previous column, “Why is the data storage industry locked in by hardware?”) that in the IT universe not all such software/hardware lock-ins last forever. The breaking of ties between Unix flavours and specific RISC processors was the stand-out example, and Linux OSes on commodity x86 servers were the agent of change to bring this about in the early part of the last decade.
The parallels of the situation and pointers to change are there in storage. We have the hardware/software lock-in, and the possibility exists to separate the software-based controller intelligence and the commodity drives underneath. We also have a very compelling motive: the potential to save a lot of money.
Let’s look at some examples of current alternatives to the existing regimes. Firstly, there’s Coraid, with its ATA over Ethernet (AoE)-based offering. Its controller software, evolved from Linux, runs commodity drives (SATA, SAS, SSD) in a scale-out fashion with Ethernet as its connect instead of Fibre Channel or TCP/IP (iSCSI).
It puts ATA commands inside Ethernet frames, so no TCP/IP overhead and no expensive host bus adapters (HBAs), and of course it’ll run over whatever bandwidth Ethernet network you want it to, so it’s potentially very fast indeed.
And it does all this at a very attractive cost; Coraid claims less than $500 per terabyte, so the sub-$10,000 array is possible here.
Then there’s Nexenta, which is a storage OS distribution based on ZFS that provides high-end enterprise NAS on commodity hardware and includes customers such as Korea Telecom, which has 20 PB installed. Nexenta claims cost savings for an installed system of 70% to 80% compared with the main vendors.
So, it’s entirely possible to come up with workable enterprise storage alternatives to the existing big-vendor regime and its software/hardware lock-in. You can go down that route with the help of companies like Coraid and Nexenta. You can even do it yourself.
As I write this, UK political pundits are abuzz with the idea of the network vs the hierarchy. It’s an idea fuelled by the Arab Spring, where Facebook- and Twitter-connected leaderless groups of activists helped mobilise masses of people to topple regimes that had lasted for decades in Middle Eastern capitals. Here in the UK the same concept is being applied to the sudden fall from positions of influence over government of Rupert Murdoch’s News International media empire.
So, could the same happen in storage? Could the network of so-called Storage 3.0 vendors and open storage devotees land some killer blows against big storage iron?
Well, maybe, but probably not too soon or too suddenly. Coraid has been floating about, doing what it does for years. It’s been five years since Robin Harris over at Storage Mojo talked of the game-changing potential of Coraid’s product. It’s hardly set the world alight since. But now it has a new round of funding, and Nexenta too is on a PR drive. In these cash-strapped times there surely has to be a potential willing market for capable enterprise storage at cut prices.
That said, there seem to be many obstacles to open source storage software. As with open source software, there is a simple lack of awareness. Add to that some variation along the lines of, “You never get sacked for buying IBM (or EMC or NetApp)” plus system integrator links to existing vendors, and it looks like the big vendors are safe in their bastions.
But then again, many people said the same thing in Tahrir Square in early January.