Free space fragmentation/write performance of SSD?

According to Diskeeper, free space fragmentation can reduce the write speed of SSD's:
http://www.diskeeperblog.com/archives/2 ... _of_f.html

This does not happen for the same reason as it does on mechanical drives, but after reading up on the topic, I still think it's possible that they are right.

SSD's are very fast when reading, and with almost zero seek time, it doesn't matter if your data is 90% fragmented. In fact, due to wear leveling even a 0% fragmented volume might be very fragmented behind the OS/filesystem layers since the data is spread out over entire range of the drive by the SSD controller.

They are also reasonably fast, but slower than conventional harddrives, when writing big chunks of data.

However, when writing small chunks of data, they are terribly slow. For example, if you want to write 32KB of data and the erase block size is 1MB, the drive has to erase a 1MB block, then rewrite whatever was in that 1MB block, together with the added 32KB of data you just saved. So your 32KB write became an erase cycle followed by a 1MB write. If there was only, say 10KB free in that 1MB block, I think it's even possible that it has to erase and rewrite two blocks to fit the data.

From what I understand, if the free space on the drive is very fragmented, even large, sequential writes become lots of small, random writes as the OS splits up the operation into several small operations to fit chunks of the file into the small holes of free space in the logical file system. This would be done by the OS, which is completely flash unaware, not by the SSD's controller, which spreads out the data intelligently to reduce wear.

Has anyone done any benchmarks comparing the write performance of the SSD with fragmented free space vs fully defragmented free space? Defraggler ((http://www.defraggler.com/) has an option to only defrag free space. It also has an option that allows some fragmentation but significantly reduces the number of moves (read/write cycles) required. In theory, running the "Defragment free space (allow fragmentation)" option once every 1-3 months should improve write performance and increase the life span of the SSD, but I haven't seen any real-world benchmarks to verify this.

 

I might try it, and do some benchmarks before and after. Your explanation sounds reasonable to me.

 

Unfortunately I accidentally overwrote the "before" data, so you will have to take my word that it didn't make an extraordinary difference. It did seem thoug, that it helped keeping the speeds up when writing files larger than 2Mb, so it may actually be useful to defrag the free space once in a while. Only took a minute or less so it's not much trouble!!