The SSD market is about to have a large shake down. So if you can hold off a while..
You may know this already but for anyone else looking at SSDs ..
SSDs - there are two forms of SSD which exist because of the two types of flash chips used to store the data:
SLC - Single-level Cell. This chip is faster (data is stored less densely which aids data transfer rates) and as a result the number of writes before the cell fails is in the order of 100,000 magnitude.
MLC - Multi-level Cell. This chip is slower than the SLC because the data (at a bit level) is stored more densely and this slows the access as the data paths are shared. The number of writes before the cell fails is in the order of 10,000.
So all SSDs have a fixed predicable lifetime based on the number of writes you perform on the drive.
Wear leveling - this is a mechanism to get around the issue of the limited number of writes that exist for each cell. In short the drive spreads the writes over the entire drive to smooth out the number of writes but also has a percentage of cells reserved that the host computer can't access to map in and replace any cells that have zero writes available. All this happens in the firmware of the drive so the host is unaware it's occurring.
Now onto the brands/makers..
OCZ Core/Solid - these are MLC and the oldest generation of drive and use the JMicron controller. It has severe compatibility issues with certain chipsets and the wear-levelling is very basic. I'd avoid these like the plague. Even SLC drives of this generation I'd avoid like the plague too (although slightly less so!)
Intel X25-E - this is an SLC drive hence it's expense for 32GB. This is the first of the next generation of drives. If obscenely fast and has a next generation approach to wear levelling. In the end when all the reserves are used the drive switches to read only so you can retrieve your data. A nice trick and shows some thought has gone into it.
This uses Intel's own controller and as a result is shot to the front of the SSD market as a result you'll pay the premium.
Intel X25-M - this is the MLC version of their SLC drive. Slightly slower but has the same next generation wear leveling firmware. As a result it's effectively captured the market and made all the other SSD manufacturers go back to the drawing board. As a result it's expensive.. but you get what you pay for in terms of quality..
OCZ Vertex - this is, as of yet, unreleased next generation MLC drive that has been made to counter Intel's devices. I fully expect it to have some of the features of the Intel drive in terms of levelling and above all far better compatibility and performance compared to the flawed previous generation. Expect to pay a premium for the early introduction.. then as competitors appear (as this is effectively a rebadged generic part) the price should drop.
This, and the like competitors, are going to enter the market and push the cost down.. hence my advice that if you can wait a bit - sit tight.
OCZ Apex, currently awaiting release, is the same Core/solid design, uses the same JMicron controller with some modifications - I'd wait until there's independent feedback to verify that the drive doesn't suffer from the same issues as the Core and Solid drives do.
The obvious thing with an SSD is to switch off drive optimisation. This is because the optimisation causes additional write wear that will reduce the lifetime of the drive and the SSD has the same flat access time over the drive. This extends to any OS running in a Virtual Machine such as windows under Parallels.
Using SSDs in RAID has the effect that blocks of writes result. So if you're writing three bytes you'll affect the full blocksize for the write. This will caused additional write wear on each drive, although if drives are in RAID0 then this may be offset slightly by distributing the write wear between the two drives.
RAID1 will cause both drives to fail at the same time (or near enough) and RAID5 will cause massive write wear due to the distributed parity blocks in addition the actual data write.
RAID is really designed for harddrives and not designed for the write/read properties of SSDs.
I upgraded my MBP to 4GB and there's a definite difference in performance. There's less use of the drive overall and for intensive tasks the machine works more efficiently due to better caching and the applications using the memory. Infact the normal low level tasks result in very little drive activity with 4GB.
For use with Parallels - I use XP/Office for work under Parallels - the machine is fast and needs 4GB to allow 1-2GB for XP. For home products using Xcode, Parallels-Ubuntu GCC compiles then 4GB again shows the 2GB a clean pair of heals purely due to the caching.
My 4GB is Crucial (DDR2 as my MBP is a Santa Rosa chipset).
I'm extremely tempted too by an SSD for the MBP. I'm looking for 120-160GB replacement for my 5400rpm 160GB drive which is good for all the activities I've used it for - although I use an external FW800 1TB drive that is miles faster for certain operations such as data storage for video processing for the robot-vision project I have.
