Although I'm busy with other projects (which use some of the following) - I've started to wonder if Database Acceleration (with the new breed of Solid State Drives/PCIe Cards and High RAM Capacity Motherboards) is enough of a task for it to be a business in its own right.

Does anyone just do this?

Do you think enough companies value the service enough (think about the hardware and power savings) to make investing in the area a good idea?

It seems to me that the severe limitations of databases (being I/O bound) are set to dissapear with well executed database schemas on High Performance IO machines.

Just looking at current technology, you can put together a Xeon server with 48GB of RAM and a very high IOPS 80/160/320GB SSD for not a lot more than a high end gaming PC costs. Granted, there are limitations that various quirks of different setups introduce - but the speed difference on such hardware (when dealing with a reasonable amount of data) is huge in comparison to any disk based system. Think about the size of caches/buffers you could have coupled with memory resident tables, then the SSD - with almost instant access and high sustained data rates - easily outperform the physical limitations of rotating disks (especially true in applications with a high degree of truly random access or access to many small pieces of data). I spoke to a big company who were adamant that disk based systems - with hundreds of disks - were better for my needs than a few raided enterprise SSDs, which was absolute nonsense given the specific requirements I have (and have since proven this to be the case with a system costing a tiny percentage of theirs outperforming their best estimate - mainly because the storage is on the PCIe bus and I need millions of small random accesses)

It seems that even recent books, that are otherwise excellent, have missed the game-changing advances in SSD performance and reliability. Am I simply wrong - are there good reasons (apart from capacity or price/GB) why this should not be a big industry for the 'changeover period'? It looks as though 1TB+ SSDs are going to be available soon (announced already but I doubt they are shipping) and there are PCIe cards already on the market that can be raided to give 1TB+ and 1 Million IOPS!

If, as I am sure there are, people are out there selling these services - how are they going down with comapnies? Are they enthusiastic or do IT departments try to hide the fact that massive sums have been spent on the wrong 'solutions' (this seems still to be the case with many top end managed hosting companies who do not support SSDs). OK, not all circumstances suit SSDs, but I'm struggling to see why - even for many large companies - the performance and capacity now available for directly-attached storage don't make this a no brainer.

Finally, are there any good books on the subject? I'd love to read more on MySQL in a HIgh PErformance IO environment.