12. "Dual-ing" Standards

The next thing I did was see how the router handled simultaneous normal 802.11b and draft 11g traffic. Again using Chariot, I set up a test using two client pairs. The reference draft 802.11g test pair for each test was a WinXP Home Dell Laptop with BuffaloTech WLI-CB-G54 Cardbus card to Win98SE Ethernet client. The second pair used a WinXP Home Compaq Laptop with various 802.11b cards to a Win98SE Ethernet client.

In an attempt to hit the major chipset types, I tested the following cards:

NETGEAR MA401 (Original Intersil PRISM II based card) ORiNOCO Gold (Agere Systems chipset) D-Link DWL-650+ (TI ACX100 chipset) NETGEAR WAB501 (Atheros 5100X dual-band chipset)

Testing Notes:

Two separate Ethernet client machines were used to avoid overloading any single CPU, All clients were on the LAN side of the router, so the routing portion of the product is not involved in the test.

In each test, I started the BuffaloTech card first, then kicked in the 802.11b card. Both cards then run simultaneously for awhile, then the BuffaloTech card stops and lets the 11b card finish by itself. As you can see from the plots below, all 11b cards performed similarly.

Figure 12: Two pair test - NETGEAR MA401
(click on the image for a full-sized view)

Figure 13: Two pair test - ORiNOCO Gold
(click on the image for a full-sized view)

Figure 14: Two pair test - D-Link DWL-650+
(click on the image for a full-sized view)

Figure 15: Two pair test - NETGEAR WAB501
(click on the image for a full-sized view)

Remember that the variation in the BuffaloTech card when it's running by itself is a known issue, and not due to any effect from the 802.11b card. The one card that seemed to be a little more 11g-friendly was the NETGEAR Atheros-based card. Figure 15 seems to show that the BuffaloTech card was able to achieve a higher average throughput when running simultaneously, but it could just be my imagination!

The 11b testing confirms what I was told during my Comdex meetings, i.e. that as soon as an 802.11b client starts sending data through an 802.11g Access Point, all the 802.11g throughput advantage is lost. The good news that the plots also show is that as long as the 11b client isn't sending data, the 11g stations will communicate at the higher rate, i.e. the association of an 802.11b client with an AP doesn't do any throughput harm.

Of course, if you're running more than one 11g card, you'd expect them to share the available bandwidth nicely. The Chariot plot in Figure 16 essentially confirms this.

Figure 16: Two pair test - Two Buffalo Tech WLI-CB-G54
(click on the image for a full-sized view)

This plot just happened to do the sharing during the downward throughput hop. It's interesting to note that whatever the mechanism is that's causing the hopping throughput, it didn't cut the throughput yet again in half when the second draft-11g client kicked in.

I, of course, ran my usual four-location throughput test, the results of which are shown in the chart at the top of this section. I also did Chariot runs, which are shown in Figure 17.

Figure 17: Four condition throughput
(click on the image for a full-sized view)

The hopping throughput problem makes the plots not very useful, but I included the screenshot so that you can see the Max, Avg, and Min throughput numbers for each run.

I also ran tests with 128bit WEP enabled, and didn't see any throughput degradation, at least none that I could separate out from the normally varying throughput.

Finally, I tried running some Turbo-mode tests, but couldn't find a Turbo setting in the Buffalo Client application. After some back-and-forth, Buffalo told me that Turbo mode is presently implemented in the AP (router) side only. The client is supposed to auto-detect the Access Point's mode, and adjust itself accordingly.

After following Buffalo's instructions and setting the router into Turbo mode, I did a Chariot run and found that, once again, I couldn't accurately see any Turbo effect due to the normal throughput variation. So the question of how much boost Turbo mode gives you, and whether it lowers effective throughput under weaker signal conditions as 802.11a's Turbo does, will have to wait for a future firmware release.

802.11g Wireless Performance Test Results

Test Conditions - WEP encryption: DISABLED - Tx Rate: Automatic - Power Save: disabled - Test Partner: Buffalo Tech WLI-CB-G54 draft 802.11g CardBus card			Firmware/Driver Versions AP f/w: 1.10 (1.14) Wireless client driver: 3.10.19.0 Wireless client f/w: No Info
Test Description	Signal Quality (%)	Transfer Rate (Mbps)	Response Time (msec)	UDP stream
				Throughput (kbps)	Lost data (%)
Client to AP - Condition 1	100	15.5 [No WEP] 15.9 [w/ WEP]	1 (avg) 1 (max)	499	0
Client to AP - Condition 2	75	16.5	1 (avg) 1 (max)	499	0
Client to AP - Condition 3	30	10.5	1 (avg) 1 (max)	499	0
Client to AP - Condition 4	45	6.9	1 (avg) 2 (max)	497	0

See details of how we test.