Why hasn't Sun been able to capitalise on its StorageTek acquisition more successfully? The answer, one answer perhaps, lies in two different approaches to storage.

Some original story text has been removed as the Freeman Reports basis for it is in doubt.

Sun and StorageTek differences

In March, 2005, Mark Canepa, then Sun's executive VP for networked storage, said: "We build storage products on a systems model, not just a storage model. Storage is an inherently threadable problem. With SPARC and Niagara technology we can have 32 threads and up to 8 cores on a single chip. It allows us to build far more computational ability into storage. Solaris, the most threaded O/S in the world, can use all these threads."

This, with hindsight, prefigures the X4500 hybrid NAS server, and Honeycomb, the 5800 content-addressed storage box for unstructured, flat-file data. This is the core engineering interest centre for storage inside Sun. Storage is a systems business for Sun's engineers, not an add-peripherals-to-systems business. Peripherals being direct-attach disk, network-attached disk, virtual tape and physical tape automation products. StorageTek is, of course, basically an add-peripherals-to-systems business.

StorageTek background pre-acquisition

Before the acquisition StorageTek's HW business was stable and secure. It had, for example, 69 percent of the mainframe virtual tape market. But the business wasn't growing and wasn't declining. StorageTek was a unique storage company in that it bridged the tape and disk array product worlds. EMC, HDS, Dot Hill, Xyratex and Engenio did disk as did 3Par, BlueArc, NetApp and others. Quantum, ADIC, Overland, SpectraLogic, Tandberg and Exabyte did tape.

IBM did both tape and disk but it was a server, systems and services company too. HP did did disk and (OEM'd) tape but it was a server and systems company. StorageTek was unique, a pure play storage vendor doing both disk and tape. But it had problems. It wasn't big in mid-range tape; mainframe tape being its bedrock and strength. It wasn't big enough in disk, not big enough to compete effectively with EMC and HDS. EMC had stolen its information lifecycle management (ILM) message and was marketing itself as the ILM company more effectively than StorageTek. NetApp was doing much more than StorageTek with filers and there were growing technologies like open systems virtual tape that needed attention.

What did StorageTek do? What could it do? Could it go it alone and build up a convincing disk message like EMC, the big gorilla on the storage block? Could it expand into mid-range tape and take on both Quantum with DLT and the LTO trio of Certance, HP and IBM? The answer it came up to both questions was 'No.' It found itself a server and systems company and got bought by Sun, a Sun very keen to decrease the EMC attach rate inside its customer base.

The acquisition's immediate after-effects

At the time of the acquisition James Whitemore was looking after storage marketing for Mark Canepa. The acquisition would end both of their Sun careers.

In tune with Canepa, Whitemore put forward the view that Sun is building storage platforms to facilitate data management. Sun is a systems company and is about combining Solaris middleware, servers and developing storage platforms to provide better data management products.

STK was absorbed with networked storage into a new Data Management Group (DMG). Thus it got the 6920, the Pirus-technology using, virtual storage controller, the OEM'd HDS TagmaStore disks and virtualising storage controller, a set of network-attached storage (NAS) products and mid-range arrays. STK bought is mainframe tape, the StreamLine product, and some mid-range automated tape products, several drive arrays, a leading mainframe virtual tape library (VTL) product, VSM, and a project to produce an open systems version of it, VSM Open.

There were three separate development/product organisations in DMG: Disk; tape; and software, together with a separate marketing organisation headed by Whitemore.

DMG inherited two Sun systems storage projects: the hybrid server/NAS Thumper, now released as the X4500; and Honeycomb, a content-addressed storage box for unstructured, flat-file data, now termed the StorageTek 5800.

There were two approaches to engineering storage products here. One was as 'storage' products, peripherals that held data and attached to servers either directly or via a network. The second is as systems-based storage products, ones with servers embedded, running Solaris and storage applications code to do something more than just provide access to raw file, block or backup set data.

In Spring, 2006, Sun CEO Scott McNealy resigned to become chairman, and favourite son Jonathan Schwartz became the new Sun CEO. This change resulted in other changes down the road.

The DMG set up demonstrated two running product problems: VSM Open and positioning the 6920 against the OEM'd HDS equivalent. It didn't solve either. In May 2006, DMG mark II appeared. David Yen replaced Mark Canepa as Sun's storage head. Canepa left and, after a few months, resurfaced as CEO of networking company Extreme. Yen re-organised DMG, combining all three divisions and Whitemore's marketing group into one operation. The ex-StorageTek and now DMG tape unit head, Nigel Dessau, was given the marketing remit. With no job anymore, Whitemore left Sun and hasn't been heard of since.

Over the next few months Yen was rumoured to be unhappy in his role. It was mooted that the 6920 would be sold, even that DMG would be sold. In the event the 6920 was put into end-of-life mode with HDS providing maintenance and support and the virtualising networked storage controller product line needed by Sun.

What seemed to have happened is that Sun people put DMG storage products into one of two camps. They could be products that Sun could transform through Sun Fire HW and Solaris and system SW injections into Sun systems storage products, like the X4500 and 5800 - or they could not, like networked storage controllers or disk drive arrays. Yen canned the 6920. It was a storage peripheral and had no real potential to be developed into Sun system storage aka data management box.

The VSM Open product did fit the DMG mould but the development project had stalled. Sun had decided to base it on Solaris, in tune with Sun's idea of building storage on a systems model. The project crashed and burned and Sun had to turn to FalconStore virtual tape software to build VTL Plus. This was, in effect, described as an interim product with the real high-end open VTL being a promised VTL Enterprise, based on a Sun Fire server, Solaris, VMS Open assets and also FalconStor assets.

Sun's system approach to storage was developed with the idea of having an open source storage system stack, embracing Solaris, NFS, ZFS - Sun's new all-singing, all-dancing file system, CIFS (in the future) and more so that storage product companies could use this freely available storage platform software and build their own storage products on top of it.

The hope was that this would generate Sun hardware and Sun service/support revenues in its wake. In this scheme disks appeared to become JBODs (just bunches of disks) with the free storage software platform and cheap commodity server hardware doing the job of turning disks into reliable drive arrays with some kind of storage application personality.

Tape continuity but not DMG continuity

This open systems storage SW/HW stack doesn't appear to do much for the STK side of the DMG business. There hasn't been any mention of a Sun Fire/Solaris/ZFS-based StreamLine tape library. That would have alarmed corporate StreamLine customers. They just wanted Sun's StreamLine and tape roadmap adhered to. And so it was. The 500GB T10000, the next-generation in StorageTek's 9940 line was launched. The next generation is on track to a 1TB capacity.

Tape seems inherently unsuitable soil for Sun's system storage plants to take root in. It's stable and mature with steady and predictable roadmaps and no competing technology, especially at the mainframe level, for the job it does best. The Freeman Reports expect tape library revenues to rise to $2.15 billion in 2012. All Sun has to do is deliver what the market wants and it should be okay.

And that was Yen's time at the DMG helm over. DMG mark III was introduced in March, 2007, less than a year after mark II. Yen moved to head up a new Microelectronics unit in Sun. Sun looked into the STK executive ranks and promoted Jon Benson, then VP tape engineering - and holder of StreamLine patents, to run Sun Storage. This was slimmed down from DMG as the NAS, X4500 and 5800 products and associated engineers were moved into Sun's systems unit run by John Fowler.

Thus was the split between the two types of storage product institutionalised. Sun storage is basically the StorageTek operation. Fowler has the true data management products.

Declining storage revenues

Recent Sun financial reports demonstrate storage is stuck in a kind of revenue limbo. There have been slight declines in revenue for both Q2 and Q3 fy 2007 on a year-on-year basis. When discussing the Q3 results, 0.2 percent down, Schwartz said: "Our storage business showed pockets of strength but wasn't overall as strong as I would have liked. Our archive or tape business is growing in what has traditionally been a weak quarter but our disk business is seeing continued challenges as we transition to a Solaris platform." He mentioned design wins for Thumper (X4500) and said it was central to Sun's strategy.

We might take the view that Sun has taken its eye off the 'bog standard' drive array/NAS ball and is pinning its hopes to NAS products with an application personality and embedded server/SW processing capability like the X4500 and 5800. There is little scope for (profitable) innovation in ordinary drive arrays. It's not that Sun is ignoring them completely, having introduced impressive new low-end arrays using SAS disks, but repetition of the Solaris, open systems storage stack message indicates which ball its eye is on.

What are we to make of this? On a favourable reading the traditional StorageTek storage business is stuck in revenue limbo facing continued decline in disk revenues and potential growth in tape library sales, but not in the mid-range where HP, IBM and Quantum, all heavy LTO product vendors, are growing, according to Freeman. Sun's system storage business is not yet delivering the revenue goods; it's seeing 'continued challenges.'

On an unfavourable reading the StorageTek traditional storage business is in trouble. Its disk revenues are poor and its tape library market share under attack from HP, IBM and Quantum. Super-NAS vendors like Acopia, BlueArc and Isilon can grow strongly with NetApp and EMC dominating the mid-range NAS space, leaving only niches for Sun.

I think a fair conclusion might be that the Sun StorageTek acquisition is still not performing quite as hoped for. However, it is far too early to go any farther and suggest that, if nothing much improves, it is heading, like the earlier Pirus acquisition, towards failure. That is an unwarranted conclusion.