Hey Lars,
Totally agree with you - Fusion-IO improves one specific piece of the HANA appliance. It does this for all single node appliances and IBM scale-out. All other vendor scale-out use SAN solutions for all data.
The redo logs are stored on Fusion-IO. So, Fusion-IO specifically improves the performance of loading data into SAP HANA whilst retaining the ACID properties of the database.
Since you can write at around 1GB/sec to a Fusion-IO card, HANA can load data into memory at up to 1GB/sec. Note that this also means that there is a theoretical improvement in replaying redo logs into memory on startup, but I've not seen that to be I/O bound.
Like you say, Fusion-IO does NOT (in any HANA appliance):
- Improve query performance
- Improve startup time (apart form redo logs)
- Store any data apart from logs
In many other database solutions, you would store your data on Fusion-IO so you improve the performance of cache misses, but this is not how HANA uses Fusion-IO.
John