Axiom to boot out NetApp at Oracle
El Reg peers into crystal ball of storage
Analysis It's surely axiomatic: Oracle will replace the NetApp E Series-sourced 6000 line of storage arrays with the Pillar Axiom.
Oracle dislikes OEM supply deals, seeing no good reason to send part of a customer's purchase money to another supplier. The company has bought the Pillar Data startup, which was funded by Oracle CEO Larry Ellison's Tako Ventures, in a zero-cash deal. Oracle now has its own unified storage – a block-storage array – in the form of the Pillar Axiom.
Pillar failed to establish standalone traction, competing with other arrays from EMC, HP, HDS, IBM, NetApp and others, and the company was rescued from a slow death by Oracle. There was only one product – the Axiom 600 – presumably as revenues were insufficient to fund continued supply and development of any others.
In his last blog, Pillar's CEO Mike Workman wrote: "We’ll be working to build products and integrate them into the Oracle stack in a way that will take everyone’s breath away. As the story evolves, we will see the true benefits of engineering the hardware and software, together."
It's now two months on, very early days indeed, but El Reg would hazard a guess that his team at Oracle, where he is SVP for storage solutions, not products we note, is doing just that: integrating the Axiom array with Oracle database software and providing a block-access alternative to Oracle's filers, the ZFS Storage 7000 line.
NetApp is happy that the supply contract with Oracle, which it inherited when it bought Engenio, will continue. But... why does Oracle need two block-storage array product lines? The 6000 line was never positioned as big data buckets or HPC data vaults, the direction NetApp marketing is going with the products.
It is conceivable that Oracle could push them as such, but it wants to build integrated hardware and software stacks using its own IP. The Axiom product fits in with that. The NetApp product does not.
Consider the Axiom's design: it ain't a traditional dual controller array, having Storage Bricks to hold the data, Slammers to control the I/O, and a Pilot to manage the shop via policy control. You can have NAS, SAN (FC) or iSCSI Slammers, and up to four of them in a 600. This is inherently more scalable than a dual controller array and it can be scaled up or scaled down.
What's in a name? A number.
There are three Oracle Sun Storage 6000 products: 6180 (to 112 drives), 6580 (256 drives), and 6780 (448 drives). To replace them and provide a classic three-box line-up of entry-level, mid-range, and high-end, Oracle would need three Axiom sub-brands. It has a naming problem too of course, since Axiom 6000 and Sun 6000 are incompatible. You wouldn't expect Oracle to call the next-gen Axiom a 6200 because of that. The 7000 numbers are taken by ZFS and a 5000 line would imply something smaller than the 6000.
How about the 8000? Oracle might be looking at something to overlap EMC's VNX line and go into VMAX territory. We could envisage an 8200, 8600, and 8800 if we wanted, and imagine the current Axiom with a brain transplant to the latest X86 processors, plus, maybe, an InfiniBand link to interconnect Slammers and Storage Bricks. That would boost the array's internal horsepower and bandwidth.
Tweak it with ZFS Storage 7000-style read and write flash caches in the Slammers, plus a healthy slug of SSD storage, and you would have fast and flash-boosted CPUs, shipping data over fast links from Storage Bricks with a flash tier, which would seem a good fast basic recipe into which to bake Oracle software.
If this were happening, and Oracle-stroke-Pillar is keeping schtum, then we would expect a Q1 2012 indication at the earliest. This analysis by the way is 100 per cent El Reg's work. We haven't received any nods or spin from insiders or anyone else. ®
Sponsored: Benefits from the lessons learned in HPC