Я читал, что самая большая слабость твердотельных накопителей - это срок их службы на основе R / W-циклов . Я работаю с большим количеством (в сотнях) - обычно очень маленьких - файлов много раз в час, и повышенная скорость чтения твердотельного накопителя является большой причиной, побуждающей меня приобрести его. (Я работаю с системами TeX, которые печально известны своим количеством файлов).
Я помещаю чрезмерную нагрузку на диск, сохраняя эти файлы на нем? Есть ли что-то интеллектуальное OS X (Mavericks, в моем случае) делает закулисные для консолидации чтения-записи и, если нет, я могу что-нибудь сделать, чтобы повлиять на это?
The primary point to keep in mind is the amount of data that is getting written and deleted per unit time. If that data is small, its not going to be an issue. If that data is large, it will be.
There is nothing that Mavricks (or any other OS for that matter) can really do about this without lying to you about what is one the disk (i.e. if you used a RAMdisk for this, you would not have this issue, but also you would not be storing your files to disk)
As I understand it, SSD drives have firmware drivers which perform load leveling algorithms. In fact, there really isn't any defragmentation on an SSD drive, like Norton Utilities used to provide. From OS theory 101, there is often a background task which is used to defragment disk drives, especially on xNIX systems; I'd assume that on a Mavericks SSD system, such processes are never launched. Analogously, if a secure erase is desired on an SSD drive, I believe that isn't really possible from the Mavericks or any other xNIX kernel or application level, because the drivers in the SSD unit perform their own remapping of writes to perform least used write block next algorithms to load level the block drive writes to extend the drive's life. Finally, if an SSD addressable memory unit becomes "bad", I think the theory is that the SSD driver simply maps out that addressable unit as bad, and the effective size of the drive becomes a bit smaller. So, with that in mind, given that Mavericks and SSD based Mac's usually come with a lot of RAM, I used to look for disk thrashing conditions (because you cannot see them or hear them if they are SSD's) a la the Activity Monitor, because such conditions are like race disk swapping conditions, and I'm sure such an event could feasibly destroy an SSD in no time at all if an OS were to let that occur, but given that the load leveling is "smart load leveling" performed by the SSD itself, I stopped being concerned as much with examining Maverick's notion of thrashing, which actually has become harder to deduce given the current dummied down Activity Monitor in Mavericks.