Fun with Throughput
The draft spec leaves manufacturers a good amount of performance-tweaking leeway in that it does not state that the "protection" mechanism must be used at all times. This is somewhat of a contradiction since one of the advantages of 802.11g is its interoperability with 802.11b. But since the "protection" mechanism eats throughput, the standards makers didn't want to saddle all-11g WLANs with protection's overhead.
As a result, draft-11g chipmakers are experimenting with a number of methods that try to minimize the protection throughput hit, yet still be able to communicate with 802.11b stations. The simplest method is to disable the protection mechanism entirely, yielding an "11g-only" mode. This is still a work in progress, with a mix of nomenclature used to describe it.
But disabling protection entirely can cause other problems that I'll describe later. So manufacturers will most likely try to get as much throughput as possible, while still providing 802.11b protection. This seems to be Broadcom's approach in the different versions of firmware I've seen, and I expect that other vendors will follow similar approaches.
However, no matter what tricks are used, it's important to understand two key points about the effect of 802.11b protection:
KEY POINT #1: All 802.11g equipment, whether draft or released, will have lower throughput when 802.11b devices are associated (connected) to the 802.11g WLAN. The lowered throughput will be seen even when the 802.11b devices are idle.
KEY POINT #2: All 802.11g equipment, whether draft or released, will slow even further when 802.11b devices are actively transmitting or receiving data.
My testing so far with only Broadcom-based products shows that the actual throughput in both Point 1 and Point 2 cases varies from product to product. This product-to-product variation will probably also remain in released 802.11g products, but it's too early to tell for sure.
I go into the throughput reduction issue in more detail in Part 2 of this NTK, but I'll give some quick examples to give you a flavor for what's going on.
• All the following tests were taken under best-case signal conditions, with AP / router and client about five feet apart. WEP was not enabled for any tests.
• The results and conclusions presented are based on draft-802.11g products only from Buffalo Tech [see review] and Linksys [see review]. Both products use Broadcom's draft-802.11g chipset, but are designed using different versions of the draft standard.
Figure 3 shows a comparison of throughput test runs for the Linksys WAP54G access point and WPC54G CardBus card, compared with the Buffalo Technology WBRG54 and WLI-CB-G54 CardBus card. Both sets of products were the first shipping versions, with original firmware.
Figure 3: Throughput comparison - Linksys vs. Buffalo Tech - original
(click on the image for a full-sized view)
The lower trace shows Buffalo Tech's first attempt at automatically handling both draft-11g and 802.11b clients, which I dubbed "throughput hopping". I'm not sure of Buffalo Tech's exact approach, but it appears that they were periodically adjusting something in the protection algorithm that cut throughput about in half. They might have been periodically turning it off completely, but other experiments that I performed make me doubt that theory.
The upper trace appears to show that Linksys has much steadier throughput, but actually is somewhat inaccurate. It turns out that Linksys' first efforts had what they called a "warm-up" effect, which Figure 4 shows.
Figure 4: Throughput comparison - Linksys before and after "warmup"
(click on the image for a full-sized view)
The lower trace is the throughput that I got when I powered up both access point and client and then immediately ran a 1 minute throughput test. Further testing showed that throughput got steadily better over about a 20 minute period, which shown in the upper trace.
Note that these throughput effects did not prevent Linksys or Buffalo Tech client cards from working with the other's router. Both cards associated, transmitted and received data without problems from either router. The main variable was the throughput of the data transmission.
The good news, however, is that the first improvements have already been issued by Buffalo Tech and Linksys. I'll cover that in the next section.