Analysis

posted on 29 April 2008 13:36

Sun is producing an open source storage software stack which is steadily being extended to create a virtual storage operating system. The idea is, that by removing proprietary storage system software from the loop, there will be a radical decrease in the cost of building storage systems, a revolution in fact. Is this making ripples, a groudswell or a tidal wave?

For many months now Sun has been promoting its efforts to provide an entire open source storage software stack so that storage system and product builders can use this free software, layer it on top of vanilla storage hardware and provide very much less expensive network-attached storage (NAS), for example, than proprietary NAS vendors such as EMC (Celerra) and NetApp.

The main ingredients of Sun's storage stack are its Solaris O/S, the ZFS file system, and various pieces of functionality such as NFS, CIFS replication and snapshots and so forth. Sun will hope developers use this software stack to build storage products. These can be based on Sun storage hardware, such as its hybrid server/storage X4500 box, or on anybody else's hardware.

Sun makes money on service contracts with the developers and on hardware revenues if the developers use, or recommend their customers use, Sun hardware.

So, all suppliers developing storage products can choose to use the Sun open source storage stack. Will they?

Let's take a mainstream high-end supplier like IBM or EMC. Would it introduce a storage area network (SAN) or NAS product using the Sun stack and so cut its costs?

It's pretty clear that they see Sun as competition and wouldn't do this. Such companies have significant IP in their software stacks, which they value and which differentiates, they think, their products from the herd. So, no, it's not at all likely that IBM or EMC or a NetApp or an HDS would use the Sun stack.

Would a supplier building product for them use the Sun stack?

Would Dot Hill, LSI or Xyratex use the Sun stack to lower their own production costs? The Sun stack runs on x86 hardware. Unless these guys use embedded x86 hardware in their array controllers then they cannot use the Sun stack. If they use firmware and not an embedded x86 server to control their arrays then they won't use the Sun software. If it became clear to their OEM customers that they were using Sun open source software in their controllers their OEMs might not feel good about it.

The overall assessment here is that Dot, LSI and Xyratex might use the Sun software but are unlikely to.

Would a new and relatively young storage product supplier, a Dell/EqualLogic, a Left Hand Networks, a Pillar Data, use the Sun software?

Not Pillar. It's IP partly resides in the way its system and controller software is integrated with the Pilot and Slammer building blocks in its product and in the way disk drives are handled. There may be commodity hardware components in there but it is built into a novel and distinctive overall system.

Not Left Hand Networks. It's IP again is actually in the storage software stack which enables it to take over the direct-attached storage in servers it runs on and aggregate it into a SAN. It could use parts of the Sun software stack in theory but, I think, there may be a channel conflict problem here. Left Hand sells its products to SMEs and enterprises and Sun and its partners sell into the same market.

Even though the Sun software is open source the fact remains that Sun, as a storage system product supplier, competes with Left Hand Networks and similar companies. For Sun to also supply them with storage system software, via open source, sets the scene for perceived channel conflict.

The overall assessment here is that storage companies like 3PAR, Atrato, Dell/EqualLogic, Exagrid, ExaNet, Left Hand Networks, Pillar, Xiotech, etc. are again unlikely to use the Sun storage software stack because they already have distinctive and differentiating IP in that area, they don't use x86 hardware, and/or they would view it as engendering potential channel conflict.

A supplier offering a storage application using a storage operating system, such as business intelligence and data warehousing supplier GreenPlum, could and does use the Sun software and hardware (X4500) to compete with other established data warehousing/business intelligence suppliers like Teradata and Netezza. It's IP is above the storage O/S. As long as it and the hardware beneath it deliver great and cost-effective performance for its DW/BI application then that is good enough.

A vanilla white box storage supplier could readily use the Sun storage software as it would add significant value to products at a very affordable cost and no channel conflict would ensue as a result.

It may be the case that suppliers like these plus a raft of developers of systems that offer their IP above a storage O/S will more naturally gravitate to the open source Sun storage SW stack; companies like GreenPlum.

End user customers that are already using open source Sun products, such as MySQL, will look more favourably on storage products using the Sun open source storage software. Customers that don't - won't.

For there to be the evolution of having a single main storage software stack used, much as there is a single main server hardware design- x86 - there needs to be a takeover of the world of storage system builders by open source software. Sun would say that the use of such open source storage software on commodity server hardware will provide equivalent performance and functionality to dedicated storage hardware and software at much greater cost.

Its the Ethernet economics argument in networking, the x86 argument in processing, both analogies used to say storage will inevitably, inexorably go the same way.

Sun may be right but it may also be wrong. 

For Sun, supplying open source storage software is a way to enable its channel and open source developers to better compete with proprietary storage system vendors, a way of nullifying their proprietary value bases. In a way its a long tail play.

Sun might hope that a large, large number of small open source storage system suppliers offering products at remarkable prices will comprehensively undercut the revenue bases of the few established and proprietary storage vendors and bring them crashing down.

Sun wants its open source storage software stack to be a disruptive technology. So far it has not been. Whether what we are seeing in Sun's open source storage developer ecosystem are mere ripples, a growing ground swell or a tidal wave is a matter of interpretation for now. Other Linux-based software attacks on proprietary bastions have been overwhelmingly successful; think Apache. This one could be too, but it will be a long haul.

^{[Chris Mellor.]}