Given a relatively high-end unit such as the Yellow Machine, I had hoped to see significant performance increases over inexpensive consumer-level NAS devices. To check, I measured the read and write performance using the iozone tool as described on this page. All tests were run under Windows XP SP2 on a Dell Inspiron 1000 laptop with 384 MB of RAM installed.
NOTE! How fast a computer can read or write data to a drive depends on many factors specific to the system running the test, so this test may not represent actual performance you'd see on your own system. The maximum theoretical data transfer rate one would expect to see on a 100Mbit network is around 12,000 kBps, so any values that exceed that number appear as a result of caching behavior, not network speed.
Figures 11 and 12 show read and write performance for a range of file and record sizes. Peak (cached) write performance is comparable to many other drives I've tested, as is the peak read performance, which doesn't have the advantage of caching.
To put these results into perspective, the plots shown in Figures 13 and 14 include data taken with the same iozone computer from previous reviews:
- Buffalo Technology Kuro Box
- Maxtor Simple Share
- Iomega NAS 200d StorCenter Pro
- LaCie Ethernet Disk mini
- Synology DS-101 Disk Station
The comparative plots shown in Figures 13 and 14 are taken with a 128MByte file size, which is large enough to bypass OS caching effects and show the real performance of the NAS devices.
You can see that the Yellow Machine fell near the bottom of the pack for read and about in the middle for write - a surprising result for such a costly product. In fact, the $160 Bring-Your-Own-Drive (BYOD) Kuro Box bested the $1,300 Yellow Box in both read and write performance. To be fair, some of this can be attributed to the Yellow Machine's RAID 5 overhead while none of the compared products incorporate RAID. But again, given the product's lofty price tag, it would be reasonable to expect better performance than products costing almost a tenth as much!