Disk I/O thoughts?

Question

I recently specced a server that would host an essentially read-only database used for web searches (so query response time is critical, and there's no transactional activity at all). The entire contents of this database are re-creatable in about 2 hours, via a bit of Oracle Text re-indexing and running a table population procedure that pulls data over a database link (and the linked database is properly archived and backed up etc), so it will be running in noarchivelog mode. However, because an outage would still be a bit inconvenient, it will run on mirrored disks. Currently, the tables and indexes in question use a 12GB cache, but I can tell we need a bit more than this -and an extra 1GB or so per month on top of that, because of data growth rates.
So, I specced a 24GB RAM server, expandable over time to 96GB. Lots of growth there. But because Oracle Text always seems to hit disk a little bit no matter how much RAM is available, I decided to spec 2 solid state hard drives in a mirrored configuration. Each disk is 120GB (sufficient for a long time to come) and has a 260MB/sec read and write speed, individually. The mirrored pair should therefore be able to achieve 260MB/sec, at least.
Unfortunately, the supplier now states that no 2.5" SSDDs will work reliably in the server (something to do with the interfaces not fitting properly). He's proposing instead that we fit the server with 4 Western Digital 10000rpm velociraptor disks, in a RAID 1+0 configuration. Each disk, individually, has a quoted throughput of 120MB/sec, so a two-disk strip 0 should (I suppose) achieve 240MB/sec, not far short of the single SSDD's throughput. The other two disks are there to mirror them, of course. (I know the difference between RAID1+0 and RAID0+1... that's not what I'm asking about here).
Which configuration would you go for? A RAID1 of a 260MB/sec SSDD? Or a RAID0 (essentially, although with a mirror backup) of two 120MB/sec traditional HDDs, assuming the hardware costs were broadly comparable in each case. (Actually, since response time is absolutely critical, I have been told that hardware costs are not a limiting factor in this purchase, so that really doesn't come into it at all).
Obviously, if the supplier says the SSDDs don't work properly, though, I don't exactly have a choice here. But I don't want to agree to get the velociraptors if they're going to be very much also-rans in the performance stakes. I'd appreciate any thoughts on which of the two storage solutions seems more attractive to you, which would provide fastest response times to web-site searches, and so on. I cannot vary the two specs by much, though: it's really only one or the other, with minor tweaking to the config possible if it seems desirable.
If it makes any difference, we'd be using OEL5.4, Oracle 11gR2, 64-bit.
I realise it's an open-ended question. Any thoughts appreciated, however.

Answer 1

No, it's a burst of read I/O that always happens whenever a CONTAINS query is submitted. It's ineradicable, too -doesn't depend on the amount of buffer cache in the slightest. Hence, I want lots of RAM to deal with the major part of handling those queries ASAP. But I want the fastest possible, though cheap and basic, physical disk platform so that the little bit of residual I/O is not a major problem.
I do understand that maximising I/O is not just about the technology used, but that's all this question is related to. I have a purchase decision to make: 2 SSDDs or 4 fast HDDs. That's the only degree of freedom in question right now. (In other words, buying 8 HDDs or 4 SSDDs is not an option: disk numbers are fixed in either case, Neither is 'using a faster I/O bus': this is a commodity (cheap!) server, and that's not something that is changeable at this time. Money is about the same either way, so that's not the constraint here... it is simply a question of whether the performance of the one would be comparable to the performance of the other, theoretically.
And yes, I understand (and did say, in fact) that if the vendor is ruling out the SSDD option for interface issues, then I don't have a lot of choice (well, I have the choice of cancelling the order, I suppose!). The question really is at that point, however, is a server with 4 HDDs (2 in a stripe) going to do disk I/O comparably to what I would have got (theoretically) with the 2 SSDDs (one mirrored)? If so, it's a viable alternative. If performance is going to be orders of magnitude worse until I get up to 8 or 16 or more spindles, however, then cancelling the order looks an attractive alternative.
I always seem to upset people here when I ask them to focus on the specific question I've asked, rather than to open it up to all sorts of other issues (such as faster bus I/O, disk partitions, and so on). I really don't intend to do so, and I hope I'm not doing so on this occasion with you, either. I suppose what I'm asking, putting it bluntly, is: will 4 HDDs in a RAID1+0 give me about the same sort of I/O throughput as 2 SSDDs in a RAID1 configuration, ideally and theoretically speaking? That's really the only thing I'm asking about. If it will, theoretically, then I'll buy the 4 HDDs. If not, I won't: it comes down to that for me, basically.
I do appreciate you taking the time to answer, though!