booting is not necessary, I need these drives for scratch disks only.
Then why do you need a 50?
I ask, since it's temporary data, a striped set will make more sense. The reasoning behind this, is threefold:
- if disk dies, you replace it and don't worry about restoring the data as it was temporary to begin with (any work being processed at the time of the failure has to be re-done anyway).
- it's faster than a 10, 5 or 50 for the same member count (or any other level for that matter).
- it's the cheapest way to go for sequential performance if redundancy isn't necessary (which is the case for temp data).
alright lets forget that card for a minute and focus on eSata, which I already have.
with mac osx, I can create a raid array regardless of the card, I have already done this creating a very fast raid 0 array on three 1TB eSata drives, however before I upgrade I wanted to do something more stable, no one really answered this main concern if I can nest hardware raid enclosures this way, nesting raid5s would seem to me to get the best speed with redundancy, however I may not be able to do this anyway due to the cost.
Yes, you can do a software nest of hardware based levels (i.e. 2x 5's on a hardware controller, then stripe via Disk Utility).
That said however, I don't see a need for redundancy for scratch (see above).
You could even consider a small, fast SSD for scratch (= low cost), as they can match a 2 - 3 drive mechanical stripe set for sequential performance.
And to top it all off, if you've enough memory, there shouldn't be much, if any need, to actually page data to the scratch volume in the first place (if you don't recall, it was a low-cost, stop-gap solution for insufficient memory back when RAM was very expensive). Photoshop does search for a scratch volume each time it's launched, but if you've enough memory in the system, it won't actually need to use it.
Which is faster? Raid 10 or Raid 5? then another question, which is faster Raid 10 in 1 enclosure? or two Raid1 enclosures on a software raid0 array?(if that is possible?)
First off, software implementations are usually slower than their hardware counterparts. Particularly for redundant levels, as it's not as detailed in terms of being able to leverage the parallelism for performance (algorithm design used for it's implementation).
Second, there are software based RAID 5 implementations, which you should avoid like the proverbial plague. The reason, is software implementations cannot deal with the write hole issue (you can find out what this is in Wiki). That solution requires hardware, which is why you need a proper hardware RAID controller for parity based levels.
Now in regards to which level is faster, RAID 5 has surpassed 10 about 5 or so years ago for both sequential and random access throughputs (10 used to be the rule for relational databases, but 5 has taken over since ~ 2006, as RAID cards became faster).
As per which way would be faster in creating a 10, it would depend on the enclosures used, particularly the RAID 1 boxes. Some have simple hardware controllers, while others are just software. Either way however, there won't be much difference if the drives used are identical, and they're run over the same eSATA card (keeping all other things equal, otherwise performance data will also reflect variances between disks used as well as the SATA controller).
To do raid 5 you'll need either additional software (often provided with the card) or hardware raid.
The simplest advice with a software based RAID 5 = do not go there, as you will get burnt (eventually will see corrupted writes due to the write hole).
