Log in

View Full Version : Some cool FTP/Wi-Fi/WZC benchmarks


Menneisyys
12-04-2004, 08:43 AM
I've originally posted this to firstloox (http://www.firstloox.org/forums/showthread.php?t=2562). Because the subject can be of much brodader/generic interest (WZC, the vast performance difference between various FTP clients etc.), I post it here too.

Please note that I've done all the benchmarks on my Fujitsu-Siemens Pocket Loox 720. The figures are likely to be different on other machines; still, the general conclusions may be the same.

I’ve run some new tests under different conditions to measure:

1, the transfer speed of all the major Pocket PC FTP clients (Resco Explorer 5+’s FTP plug-in (http://www.resco-net.com/explorer.asp), CedeFTP 2.5.3 (http://www.noisette-software.com/products/pocketpc/CedeFTP/) and vxFtp (http://www.cam.com/vxftp.html)) to the main memory, to flash cards and to a USB hosted hard disk.

This question is of extreme importance when you have a very fast (e.g., academic or corporate, not just a slow home DSL/cable) connection and want to transfer files of some 10 Mbytes. For example, you’re downloading mp3’s/DivX’s from your online repository to watch/show your friends “on the field”, where the fastest download speed is of extreme importance because your friends are eagerly waiting for the download to finish.

As it turned out, vxFtp is by far the fastest client when transferring to the main memory. (See the importance of the size of the main memory? iPAQ hx4700 users, anyone? :E ) It’s at least 2 times faster than both Resco and CedeFTP.

When it comes to download straight to a flash card or the USB host, CedeFTP becomes the fastest of all. However, vxFtp isn’t very slow in this respect either – it’s “only” 2 times slower when writing straight to SD (while, for example, Resco is an order of magnitude slower).

Resco Explorer’s FTP plug-in is a total loser in both downloading to main memory and, especially, to storage cards.

2, the effect of weak Wi-Fi signal on the Wi-Fi speed.

As it turns out, it has little effect because it’s not Wi-Fi’s actual speed that is the bottleneck on a PPC, unlike with much more powerful desktop/notebook computers. As can be seen from the results, the maximal throughput doesn’t really descend even under harsh conditions. Of course, connection _stability_ is a totally different question.

3, the effect of enabling/disabling Wireless Zero Configuration (WZC). As WM2003(SE)’s WZC is continuously trying to find other AP’s even when bound to a specific peer, you can’t achieve the same speed with WZC enabled than with WZC disabled.

As it turns out, when you enable WZC, the performance penalty is some 8% at around 300 kbyte/s (it was the only case that I’ve checked – although, several times just to be sure. The performance penalty may be smaller at slower speeds – e.g., when the Access Point is sharing a slow connection). It’s not that much – still, the results are worth considering; WZC is a real pain in the neck anyway, not just because of its negative effect on the transfer speed.

Comments are welcome!

Numeric results:

Weak (about -85 dBm), encrypted (128 bit WEP) signal:

vxFtp:
to main memory: 285 kbyte/s
to Ridata 40* CF: 71 kbyte/s
to SanDisk 256M SD: 50 kbyte/s

Resco:
to main memory: 136 kbyte/s
to Ridata 40* CF: 14 kbyte/s
to SanDisk 256M SD: 8 kbyte/s

CedeFTP:
to main memory: 150 kbyte/s
to Ridata 40* CF: 90 kbyte/s


Average (about -63 dBm), encrypted (128 bit WEP) signal:
CedeFTP:
to main memory: 154 kbyte/s
to SanDisk 256M SD: 101 kbyte/s
to USB hosted MamboX P-353: 72.2 kbyte/s

vxFtp:
to main memory, with WZC disabled: 310 kbyte/s
to main memory, with WZC enabled: 287 kbyte/s
to SanDisk 256M SD: 52 kbyte/s
to USB hosted MamboX P-353: 60.5 kbyte/s

Menneisyys
12-09-2004, 07:59 PM
I've measured the same with my iPAQ 2210 + Asus CF Wi-Fi combo.

vxFtp, disabled WZC:
to main memory, 266 kbytes/sec
to SD, 87 kbytes/sec

So, vxFtp main memory/flash card transfers are equally fast/slow on all platforms.

Travel
02-23-2008, 03:33 AM
I have applied it to my site but I can not messure the results? I have a megaphone (http://www.bullhornmegaphone.com/) system.