File Sharing Performance
Since I'll be testing more NAS devices, I figured it was time to come up with a better performance method than the Windows drag-and-drop that I'd been using. After evaluating a number of tools, I settled on IOzone, which I'll be using until something better comes along. The full testing setup and methodology are described on this page, so I'll just concentrate on the results here.
IOzone runs a series of read and write tests using different record and file sizes. But to keep the comparison plots simple, I use only the results from writing and reading a 4MByte file - which are representitive of performance with other file sizes. Note that the plots' Y axis has a logarithmic scale, for reasons you'll see shortly.
Figure 15: Write performance comparison - 4MByte file size
Figure 15 shows that the ASAP's write performance is similar to writing to a shared Win98SE folder - around 1.5MBytes/sec - but with very different results for the BOSS! I ran this test three times and each time the BOSS's write performance for record sizes smaller than 512 kBytes was a phenomenal 43 - 65MBytes/sec! My only explanation for this is that there must be a significant difference in caching between the ASAP's Hitachi / IBM and BOSS' Samsung drives.
Update 3/5/2004 - Helpful reader Roy Franz pointed out my math error in that it's not possible to get the 43-64MBytes/sec I measured from a 100Mbps Ethernet connection. So I can only conclude that IOzone's results for the BOSS write tests are incorrect and should not be used.
Update 4/15/2004 - Don Capps, the current maintainer of iozone set me straight with some very helpful and useful information about iozone. The gist of it is that the >50MByte/sec speed that I got from the BOSS' test run is due to Windows file caching effects and not an IOzone error.
Figure 16: Read performance comparison - 4MByte file size
Read performance (Figure 16) was more consistent, with both the ASAP and BOSS ranging between 1.1 - 1.8MBytes/sec. This was slightly better than a Win98SE shared folder, but not as good as the shared WinXP folder, which averaged almost 9MBytes/sec.
Since this is my first use of a new NAS evaluation tool, I can't compare these results to any previously tested NAS devices. But suffice it to say that if you've been sharing files using shared folders on a Windows machine, in most cases you probably won't notice any performance difference - for better or worse - when using the ASAP or BOSS.