Is Sun's open source strategy going to work?
Sun believes that what has happened to servers over the past decade is going to happen with storage. We have moved, and are still moving, from proprietary minicomputers to industry-standard servers running cheap and cheerful Linux. Sun's thinking is that proprietary, restrictive, inflexible and high-priced storage arrays are going the same way and will yield to those using industry-standard and commodity components with open source software.
Let's take a look at a fairly average mid-range and modular storage array, the sort of product that Sun's open storage strategy aims to displace.
It is a dual RAID controller system with 4Gbps Fibre Channel connectivity and five to 224 drives, either Fibre Channel (performance) or SATA (capacity). In its SATA guise it holds up to 224TB. The product is presented as a block storage device suitable for SAN use and its price starts at $217,995 and heads on up. This is a quarter of a million dollar box. It is the Sun StorageTek 6540.
Now hear this. On November 12th, Sun launched its Amber Road or 7000 series of storage arrays with the 7410 being a clusterable system. It can have up to 228 SATA drives, meaning 228 TB, tens of gigabytes of read-biased and write-biased (read optimised) solid state storage, and network and tape device connectivity. The 7410 has a controller, its own embedded server and four quad-core AMD processors. This runs the Open Solaris operating system and ZFS, the file system that makes RAID controllers redundant. The FishWorks software layered on that includes replication and DTrace analytics software.
The 7410, like the other 7000 products, includes block-level access via iSCSI plus CIFS and NFS file-level access. It also includes snapshot functionality. In other words, we can say that, generally, it does pretty much the same as the 6540 plus a whole lot more and is priced a whole lot less. The starting price is $89,490 for a clustered 2-server node 12 TB system and $57,490 for a single node 12 TB system.
So, shall I pay a quarter of a million dollars or $90,000? That's a hard choice. . .not.
There has to be that sort of cost saving to prompt you to take on an open storage box, like a Linux server, because you have to do more . . .hence the DTrace analytic software. Is this cost-saving enough to get potential customers interested and is this a sensible strategy for Sun?
It will surely get some customers interested, but how many no one knows.
Sun's pitch will be that this is only the first set of products. There will be others and it took several years for Linux servers to become the force they are today. There were and are several Linux distributions whereas there is only one open storage distribution. The sense I get from this is that there will be no open storage revolution, rather a steady build-up of sales.
It would take considerable market penetration by Sun's 7000 products to prompt other storage hardware vendors to follow suit and introduce their own open storage line. IBM could do this if it wanted, being a champion of Linux and open source software. Both Novell (SuSE) and RedHat might consider adding the storage system software functionality that Sun is shipping with their own Linux distributions. If various storage hardware suppliers took this up, then the stage might be better set for a true open storage movement. Until then it is Sun's gig and Sun's alone.
For Sun this open storage strategy is a two-edged sword. It will provide terrific price competition for every other storage supplier with proprietary modular arrays, but also for Sun's own products. The 7410 is aimed at displacing the 6540 and the 7000 range and undercuts Sun's block-level 6000 storage array line. Not so its file-level 5200/5300 products. A two-node 5320 NAS cluster scalable up to 420 TB starts from $64,634.
But the long-term logic of open source storage software and the commodity hardware concept is remorseless and is aimed at replacing proprietary networked storage arrays, including Sun's own.
Sun will have to sell three or more 7410s than it does 6540s to make the same profit. Can its $3 billion cash mountain provide the cushion needed as it transitions its own storage offering from proprietary and high-priced to open and much lower-priced products?
Pile it high and sell it cheap. It has worked in retail. It's working in servers. Can't say it's working in networking. Where is the Linux open source alternative to Cisco's proprietary operating system? Will it work in storage?
Come back in five years time and we'll see.
Chris Mellor is storage editor for The Register.