Wireless HygieneBy CWNP On 03/24/2010 - 22 Comments
If there’s one term that is central to wireless networking, it is this: half-duplex. I’ve heard it over and over in the past few weeks in conversations with vendors and consultants who continue to wonder why some folks still don’t understand the basics of WLAN design. As a foretaste to our real solution, here are a few broad and specific questions and answers for basic wireless hygiene that should help our friends with underperforming networks.
Is your network utilization through the roof? If so, get out your network analysis tools, dig into the details, and find the culprit. Are you broadcasting 8 SSIDs per AP? I’ve heard some horror stories lately where large enterprises are broadcasting far too many SSIDs from each AP. In one example, the CEO even had his own personal SSID. Channel utilization was above 50% with zero data traffic. Ug. One beacon frame (especially an 802.11n beacon) every 100 ms may not seem like much, but multiply that by 8, and then add neighboring APs on the same channel, and all of a sudden you’ve got a problem. If you’re seeing persistent performance problems but haven’t checked on your network utilization lately, what are you waiting for? Get your hands on some third-party (or your infrastructure vendor’s) tools to check it out. If it’s high, find the source and see how you can offload or minimize that traffic. Better yet, talk to your supervisor and tell him/her it’s time for 802.11n or some other redesign efforts.
Do you have too many APs? A few years ago, the question (and the problem) may have been the opposite: Do you have enough APs? Too many networks today have too many APs. Where network administrators and decision makers aren’t RF savvy, some salespeople will gladly sell 50 APs for a 50,000 sq. ft. warehouse. 1 AP per 1000 sq. ft.? If that doesn’t seem screwy to you, it should. It’s really impossible to say how many sq. ft. an AP should cover, but on average, you should see an AP about every 4,000-5,000 sq. feet. Of course, this depends upon many factors like client density, applications, RF environment, transmit power and antenna choice, etc, so it’s not a hard and fast number. There may be times for an AP per 1000 sq. ft., but these are unique cases. Check your AP density and see if APs need to be tuned down or powered off altogether. Are your channel and power settings askew? This is similar to the problem above, but may not be directly correlated. This topic could warrant a full day of instruction, but let’s stick to the basics here. Are your channel plans utilizing the full frequency space with non-overlapping channels? Are you saturating a single channel unintentionally? Are you effectively minimizing collision domains? How many APs are visible on a given channel and at what signal strength are they measured? It is good practice to do post-installation surveys and then perform periodic surveys of the RF environment thereafter. RF is dynamic, after all. For the inexperienced, infrastructure-managed power and channel assignments may be a good choice. If you’re wondering what to look for, you might start by charting signal metrics by doing a walkaround with a site surveying tool. Observe the RSSI and SNR relationships, and compare same-channel APs. You may find that power settings are too high or low, channel reuse wasn’t properly planned, or the noise floor is high throughout the facility. Of course, signal metrics are only a part of the picture. Real application data is what matters, so if you’re having problems with the application, check on your design fundamentals.
Do you require all data rates? One simple tuning measure that may help congested networks is to disable lower data rates like 1, 2, and 5.5 Mbps. Of course, this recommendation comes with a bit of caution. Some applications (and legacy PHYs) have dependencies on these rates. Meanwhile, certain driver builds or clients may have odd reactions to this type of change. When in doubt, do some limited scale testing—preferably in a lab—to see if any abnormal behavior results. The whole point here is that clients at the edges of cells may be “sticky”, which keeps them associated to an AP while at a low data rate. By pruning these clients off, network efficiency improves. When networks are properly designed, disabling lower data rates may have the intended impact of improving network wide performance. Are you utilizing the 5 GHz band? Band steering is a wonderful proprietary feature that most vendors offer to move clients into the 5 GHz band, when possible. Everyone should know by now that 2.4 GHz is crowded. Features that maximize the available frequency space are sorely needed in many networks. Check with your vendor to see if this feature is an option.
Are your legacy clients holding up faster clients? This problem is somewhat similar to the data rate issue mentioned previously. In this case, legacy clients that are restricted to lower data rates (such as 802.11b) are like the slow guy at the Chick-Fil-A drive-thru. Proprietary features like airtime fairness keep the network running optimally by managing airtime usage so faster clients get their fair share of opportunity. If your client population is highly disparate, this may be a good option. Check with your vendor to see if this feature is an option.
I know this is only a short list of potential design problems, but if many of the conversations I’ve had in the past month are any indication, these issues are fairly common. There are plenty of networks that need major overhaul. Then there are others that could use a few tweaks to get things running smoothly. It’s basic network hygiene, proactive and preventative, as well as reactive and responsive.Tagged with: Design, utilization, half-duplex