Troubleshooting Wireless LAN Installations

Solutions for Co-location Throughput Problems

As a wireless LAN installer or administrator, you really have two choices when considering access point co-location. You can accept the degraded throughput, or you can attempt a workaround. Accepting the fact that your users will not have 5 Mbps of actual throughput to the network backbone on each access point may be an acceptable scenario. First, however, you must make sure that the users connecting to the network in this situation can still be productive and that they do not actually require the full 5 Mbps of throughput. The last thing you want to be responsible for as a wireless LAN administrator is a network that does not allow the users to do their jobs or achieve the connections that they require. An administrator's second option in this case is to attempt a workaround. Below, we describe some of the alternatives to co-location problems.


Use Two Access Points

One option, which is the easiest, is to use channels 1 and 11 with only 2 access points, as illustrated in Figure 9.11. Using only these two channels will ensure that you have no overlap between channels regardless of proximity between systems, and therefore, no detrimental effect on the throughput of each access point. By way of comparison, two access points operating at the maximum capacity of 5.5 Mbps (about the best that you can expect by any access point), give you a total capacity of 11 Mbps of aggregate throughput, whereas three access points operating at approximately 4 Mbps each (degraded from the maximum due to actual channel overlap) on average yields only 12 Mbps of aggregate throughput. For an additional 1 Mbps of throughput, an administrator would have to spend the extra money to buy another access point, the time and labor to install it, and the continued burden of managing it.


In certain instances, the extra 1 Mbps of bandwidth might still be advantageous, but in a small environment, it might not be practical. Don't forget that this scenario applies only to access points located in the same physical space serving the same client base, but using different, non-overlapping channels. This configuration does not apply to channel reuse, where cells on different non-overlapping channels are alternately spread throughout an area to avoid co-channel interference.


Use 802.11a Equipment

As a second option, you could use 802.11a compliant equipment operating in the 5 GHz UNII bands. The 5 GHz UNII bands, which are each wider than the 2.4 GHz ISM band, have three usable bands, and each band allows for four non-overlapping channels. By using a mixture of 802.11b and 802.11a equipment, more systems can be co-located in the same space without fear of interference between systems. With two (or three) colocated 802.11b systems and up to 8 co-located 802.11a systems, there is the potential for an incredible amount of throughput in the same physical space. The reason that we specify 8 instead of 12 co-located access points with 802.11a is that only the lower and middle bands (with 4 non-overlapping channels each) are specified for indoor use. Therefore, indoors, where most access points are placed, there's normally only the potential for up to 8 access points using 802.11a compliant devices.

Issues with 802.11a Equipment

802.11a equipment is now available from only a few vendors, and is more expensive than equipment that uses the 2.4 GHz frequency band. However, the 5 GHz band has the advantage of many more non-overlapping channels than the 2.4 GHz band (8 vs. 3), allowing you to implement many more co-located access points. You must keep in mind that while the 2.4 GHz band allows for less expensive gear, the 2.4 GHz band is much more crowded, which means you are more likely to encounter interference from other nearby wireless LANs. Remember that 802.11a devices and 802.11b devices are incompatible. These devices do not see, hear, or communicate with one another because they utilize different frequency bands and different modulation techniques.

Troubleshooting Wireless LAN Installations

System Throughput

Throughput on a wireless LAN is based on many factors. For instance, the amount and type of interference may impact the amount of data that can be successfully transmitted. If additional security solutions are implemented, such as Wired Equivalent Privacy (WEP—discussed in depth in Chapter 10, Wireless LAN Security), then the additional overhead of encrypting and decrypting data will also cause a decrease in throughput. Using VPN tunnels will add additional overhead to a wireless LAN system in the same manner as will turning on WEP.

Greater distances between the transmitter and receiver will cause the throughput to decrease because an increase in the number of errors (bit error rate) will create a need for retransmissions. Modern spread spectrum systems are configured to make discrete jumps
to specified data rates (1, 2, 5.5, and 11 Mbps). If 11 Mbps cannot be maintained, for example, then the device will drop to 5.5 Mbps. Since the throughput is about 50% of the data rate on a wireless LAN system, changing the data rate will have a significant impact on the throughput.

Hardware limitations will also dictate the data rate. If an IEEE 802.11 device is communicating with an IEEE 802.11b device, the data rate can be no more than 2 Mbps, despite the 802.11b device’s ability to communicate at 11 Mbps. Correspondingly, the actual throughput will be less still—about 50%, or 1 Mbps. With wireless LAN hardware, another consideration must be taken into account: the amount of CPU power given to the access point. Having a slow CPU that cannot handle the full 11 Mbps data rate with128-bit WEP enabled will affect throughput.


The type of spread spectrum technology used, FHSS or DSSS, will make a difference in throughput for two specific reasons. First, the data rates for FHSS and DSSS systems are quite different. FHSS systems are typically in compliance with either the OpenAir standard and can transmit at 800 kbps or 1.6 Mbps, or the IEEE 802.11 standard, which allows them to transmit at 1 Mbps or 2 Mbps. Currently, DSSS systems comply with either the IEEE 802.11 standard or the 802.11b standard, supporting data rates of 1, 2, 5.5, & 11 Mbps. The second reason that the type of spread spectrum technology will affect throughput is that FHSS incurs the additional overhead of hop time.

Other factors limiting the throughput of a wireless LAN include proprietary data-link layer protocols, the use of fragmentation (which requires the re-assembly of packets), and packet size. Larger packets will result in greater throughput (assuming a good RF link) because the ratio of data to overhead is better.

RTS/CTS, a protocol used on some wireless LAN implementations and which is similar to the way that some serial links communicate, will create significant overhead because of the amount of handshaking that takes place during the transfer.

The number of users attempting to access the medium simultaneously will have an impact. An increase in simultaneous users will decrease the throughput each station receives from the access point.

Using PCF mode on an access point, thereby invoking polling on the wireless network, will decrease throughput. Polling causes lower throughput by introducing the extra overhead of a polling mechanism and mandatory responses from wireless stations even when no data needs to be sent by those stations.


Co-location Throughput (Theory vs. Reality)

Co-location is a common wireless LAN implementation technique that is used to provide more bandwidth and throughput to wireless users in a given area. RF theory, combined with FCC regulations, allows wireless LAN users in the United States three nonoverlapping RF channels (1, 6, and 11). These 3 channels can be used to co-locate multiple (3) access points within the same physical area using 802.11b equipment, as can be seen in Figure 9.9.


When co-locating multiple access points, it is highly recommended that you:

1. Use the same Spread Spectrum technology (either Direct Sequence or Frequency Hopping, but not both) for all access points

2. Use the same vendor for all access points

The portion of the 2.4 GHz ISM band that is useable for wireless LANs consists of 83.5 MHz. DSSS channels are 22 MHz wide, and there are 11 channels specified for use in the United States. These channels are specifically designated ranges of frequencies within the ISM band. According to the center frequency and width given to each of these channels by the FCC, only three non-overlapping channels can exist in this band. Colocation of access points using non-overlapping channels in the same physical space has advantages in implementing wireless LANs, so we will first explain what should happen when you co-locate these access points properly, and then we will explain what will happen.

Theory: What Should Happen

For purposes of simplicity in this explanation, we will assume that all access points being used in this scenario are 802.11b-compliant, 11Mbps access points. When using only one access point in a simple wireless LAN, you should experience actual throughput of somewhere between 4.5 Mbps and 5.5 Mbps. You will never see the full 11 Mbps of rated bandwidth due to the half-duplex nature of the RF radios and overhead requirements for wireless LAN protocols such as CSMA/CA.

The RF theory of 3 non-overlapping channels should allow you to setup one access point on channel 1, one access point on channel 6, and one access point on channel 11 without any overlap in these access points' RF band usages. Therefore, you should see normal throughput of approximately 5 Mbps on all co-located access points, with no adjacentchannel interference. Adjacent-channel interference would cause degradation of throughput on one or both of the other access points.

Reality: What Does Happen

What actually happens is that channel 1 and channel 6 actually do have a small amount of overlap, as do channel 6 and channel 11. Figure 9.10 illustrates this overlap. The reason for this overlap is typically that both access points are transmitting at approximately the same high output power and are located relatively close to each other. So, instead of getting normal half-duplex throughput on all access points, a detrimental effect is seen on all three. Throughput can decrease to 4 Mbps or less on all three access points or may be unevenly distributed where the access points might have 3, 4, and 5 Mbps respectively.


The portion of the theory that holds true is that adjacent channels (1, 2, 3, 4, and 5, for example) have significant overlap, to the point that using an access point on channel 1 and another on channel 3, for example, results in even lower throughput (2Mbps or less) on the two access points. In this case, in particular, a partial overlapping of channels occurs. It is typically seen that a full overlap results in better throughput for the two systems than does a partial overlap between systems.

All this discussion is not to say that you simply cannot co-locate three access points using channels 1, 6, and 11. Rather, it is to point out that when you do so, you should not expect the theory to hold completely true. You will experience degraded throughput that is significantly less than the normally expected rate of approximately 5 Mbps per access point unless care is taken to turn down the output power and spread the access points across a broader amount of physical space.