7. Performance

I generally dislike using Windows drag-and-drop file transfers to evaluate product performance because it's such a black box with large overhead that produces widely varying results. But since the 1100 is a file server and I didn't want to set up NetBench to beat on it, I decided to forge ahead with using drag-and-drop.

I used two clients connected via the 10/100 switch in my SMC Barricade main router at 100Mbps, full duplex for my tests:

733MHz Pentium III, Win98SE, 318 MB RAM, Accton EN1207D 10/100 PCI Ethernet 1 GHz Celeron, WinXP Home, 128 MB RAM, 3Com 3C905C-TX compatible integrated 10/100 Ethernet

I used two folders each containing the same 32 MB of 469 gif and jpg files of varying sizes, one for each of the client machines. I used a stopwatch to time the transfers from the time I dropped the file until the file transfer progress bar disappeared, and deleted each transferred folder and emptied the Trash after each test. Each test was run three times and I averaged the results.

I also tested file transfers using a 500MHz AMD-K6-2 running Win98SE with 252 MB of memory as a comparison "server". Table 1 tells the tale.

Test	Transfer Time - SnapServer (sec.)	Transfer Time - Win98SE "server" (sec.)
WinXP client	16	35
Win98SE client	30	27
Both simultaneous	50	36
Table 1: File Transfer Test results

The results show an almost 70% increase in transfer time (vs. the longest individual transfer) when both clients were simultaneously pulling files from the 1100. By comparison, the simultaneous transfer time using the Win98SE "server" was essentially the same as the longest single client transfer. I have to say that I was surprised by this result and re-ran the tests, but the results were essentially the same.

Since I'm always having readers tell me that I should use a *nix system for these types of tests, I fired up my system running SuSE Linux 7.3 and copied the same test folder from the 1100 using the Linux cp command. I have to say my readers are right because it took only about 4 seconds for the transfer to complete!

Since the main thing I learned from this exercise is that the OS overhead can be a significant factor in file transfer performance, I have to give Snap the benefit of the doubt and say that the 1100 should be able to adequately handle the needs of most home and small office LANs.