Figures 3 and 4 show the cached read performance of both drives.

Figure 3: Xbox NAS Read performance
(click image to enlarge)

There's essentially no difference in performance between the two, which supports the results I obtained with my file copy test.

Figure 4: PC to PC Read performance
(click image to enlarge)

Since the above tests are dominated by caching performance, we need to look beyond OS-specific effects to the raw device performance. Experience with iozone has shown that most caching effects are gone with file sizes of 128MBytes and higher. So Figures 5 and 6 show comparisons of write and read performance of the modified Xbox and shared notebook drive using a 128MByte file size and record sizes from 64 to 16384 kBytes.

Figure 5: 128MB Write performance comparison
(click image to enlarge)

Figure 5 gives a slight advantage to the notebook drive for writing smaller record sizes, but the Xbox NAS' more consistent performance wins out for record sizes above 512 kBytes.

Figure 6: 128MB Read performance comparison
(click image to enlarge)

On the other hand, Figure 6 shows that the shared PC drive has both more consistent and higher non-cached read performance than the modified Xbox. But as my file-copy test showed, this advantage may not show up in real-life usage.

This brings us to the end of our adventures with our Xbox NAS. It's been a bit of work, but well worth it. For not a lot of money, you can turn one person's dust collector into a fast, flexible networked storage device that also can be made to perform additional tricks such as handling your BitTorrent downloads. And since you now have a nice little Linux-based server, even more mods are possible.