Next, since the mini can also act as an attached USB drive, I ran the same iozone test with the device directly attached to a USB 2.0 port. In this mode, the best theoretical transfer one could possibly see would be around 58,000 kBytes/Second. Values in excess of this are once again caused by operating system caching.

Figure 10: USB mode Read performance
(click image to enlarge)

This particular test caused me a bit of head-scratching. As you can see in Figure 10, the read test showed the mini taking advantage of operating system caching, but the write test did not. This was in contrast to a test run against a "standard" USB attached drive which showed caching effects for both reads and writes with values far in excess of 58,000 kBytes/Second.

After some additional testing and consultation with LaCie, it was theorized that the difference was due to the mini's use of the FAT32 filesystem. When the mini was re-formatted with NTFS, performance increased and showed filesystem caching effects for both read and write. The downside to this "fix", is that a NTFS formatted disk becomes read-only when the mini is used in NAS mode.

Figure 11: USB mode Write performance
(click image to enlarge)