Dealing with the revolution

Wear levelling
Some flash devices, including USB keys and SSDs, use a technique known as wear-levelling to help prolong the life of the media.

Wear-levelling attempts to work around the limited write-cycle lifespan of blocks by arranging data so that writes are distributed evenly across a medium, rather than re-writing files in place. This way no single block prematurely fails due to a high concentration of erase-write cycles. However the efficiency and implementation of these algorithms is important – given the minimum erase block size, you don't want wear-levelling working against itself as it distributes data across a disk, erasing large blocks only to write small segments in the process.

Wear levelling is currently implemented in many devices by the onboard controller at a basic level, but more advanced methods can be implemented by filesystems specifically designed for SSDs, as well (see Linux and SSDs below).
Overall, however, the biggest single contributor to SSD lifespan is how frequently writes occur. And this, as well as the actual performance of an SSD, is all down to the operating system.

Windows and SSDs
Both Windows and Linux are tuned for spinning platter disks and use queuing alogrithms and I/O schedulers to optimise performance specifically for them. Great for hard drives, not so great for SSDs.

One problem is the I/O scheduler, which usually has two core tasks: keeping multiple disk intensive processes happy, and optimising throughput for a given medium.

Windows has just one scheduler, which in Vista is designed to provide prioritised I/O requests and bandwidth reservation as needed. It's also designed to use small 512-byte sector transfers, which works well with disk drives but gets in the way with SSDs – they fare better when the sector size matches the default 4k page size of the memory. Unfortunately, the 512-byte transfers are hard-coded into Windows.

All this leads to less than stellar performance (yes the benchmarks you've seen so far should be even higher), and when you add this to Windows' habit of constant drive accesses with its myriad of background processess (everything from SuperFetch to defragmenting), it's not doing any favours for SSD lifespan either. So it's no surprise that SSD maker Samsung recently announced it would work with Microsoft to help improve the situation. Another SSD manufacturer, SanDisk, also recently publicly raised the issue, but went as far as to say that what's needed are new SSD optimised controllers to "make up for the shortfalls in Windows". Ouch.

And all this says nothing of how NTFS affects the situation – a filesystem designed from the ground up for hard drives – not optimised for dealing with how SSDs fetch and cache data, or for maximising the longevity of the medium.