Database Storage
I seriously can't imagine Oracle being in the storage hardware business in the long term. I would expect Oracle to seek to recoup some of the purchase price of SUN by attempting to sell off the storage hardware business to other vendors (assuming it can find any with deep enough pockets). HP, IBM, HDS and NetApp could all be in the fray.
In terms of storage software, that's a completely different issue. A software set that would allow commodity servers and networks to provide storage functions is certainly possible. With ZFS, Linux, Solaris X86 and clustering software they have many of the building blocks. HP are positioning the PolyServe product they acquired in much the same area. However, it takes time, and I remain to be convinced that customers really want general purpose servers as storage servers, but this sort of thing could be built into an appliance-type model with resilience and scale It's goping to be difficult, and unlike databases, there is less of a technical lock-in with any one storage supplier. With Oracle it is often not
One problem that Oracle are going to have to come to terms with the is the fundamentally different Open Source licesning models of what they have acquired from SUN and the Linux-based software that they have developed. The latter is GPL, unless they GPL'd the SUN open source software then there would be big problems merging the two (hence ZFS is a problem to put onto Linux in the way that it is installed into Solaris).
As far as storage and databases is concerned, in the case of high-performance transactional storage we are in the next 2-3 years going to get a radical shift to flash-based solid state storage. Already on a cost per IOP basis flash is ahead (and cost per IOP is often the big problem on transaction databases so you have to provide far too many spindles - just see any TPC'C' benchmark). Cost be GB is still going to be a long way behind hard disk for the forseeable future, but not so far behind that of the fastest, short-stroked enterprise disks. As far as latency is concerned, then hard disk can't get anywhere near competing. Fifty microsecond or better latency is possible with flash with hardisk being 100 times longer (uncached). Whilst this is going to revolutionise many transactional databases, it is also a fact that none of the current enterprise array manufacturers has storage solutions which can make full use of the number of IOPs that an array with a significant number of flash drives (especially SLC ones) could achieve. They simply run out of processing power. Also the FC protocols, SAN/Ethernet fabric pretty well gurantee not seeing better than 500 microsecond I/O latency.
The stage is set for a revolution in transaction database performance once the storage problem is fixed. Putting SSDs into current arrays is only a partial measure. I will be interested to see what Oracle come up with as the fundamentals have changed (the fastest SSDs by far site directly on PCIe buses, not on the ends of FC SAN networks).