A typical network lifecycle can be broken down into four phases: design, installation, configuration, and maintenance. Design begins with capacity and needs assessment. This consists of determining how many users will be on the network, what applications they’ll be using, and what performance they will require, typically in terms of bandwidth and latency metrics. Based on the capacity and needs assessment, a plan is prepared, including specific equipment that will be purchased and how the equipment will be connected together. Once the equipment is purchased, the installation phase begins. In this phase, the equipment is placed into its final location. Installation may involve something as simple as mounting the equipment into a wiring rack, but for wireless networks, it often involves a more complex process of mounting equipment to walls, masts, or antenna towers. In the configuration phase, the equipment is configured to be compatible with the existing network, other new pieces of equipment, and the users’ equipment. After the equipment is configured and in use, the maintenance phase begins. This phase lasts for essentially the life of the equipment, and includes any maintenance that is necessary to keep the equipment working to its original specifications.
A structured network eases the labor associated with all four phases of a network’s lifecycle. In a structured network, the individual components interact in a manner that minimizes the amount of manual intervention required from the network administrator or designer. This might mean that the equipment automatically configures itself appropriately for the network; it might mean that they are designed in such a way that configuration isn’t necessary. The specific manner in which a network becomes structured varies. A structured network is defined not by how it becomes structured, but by the characteristics that result from its structured nature. Four main characteristics of a structured network are:
It is easy to deploy.
It is scalable.
It is extensible.
It is centrally-managed.
Four Characteristics Of a Structured Network
Ease of deployment means that the network can be installed without excessive up-front design. General guidelines can be used to go from capacity assessment to a rough inventory list for the network. For example, it might be known that a certain model of Ethernet switch can support up to 100 users with an aggregate bandwidth of 1 Gbps. Given that guideline, it should be easy to determine roughly how many switches will be required to support a certain number of users and/or a certain aggregate bandwidth. “Ease of deployment” means that the equipment should be flexible or “smart” enough that the designer should be able to focus on large-scale, architectural decisions without worrying too much about low-level details.
Scalability means that additional areas of connectivity can be added or areas of connectivity can be removed without requiring significant changes to the other areas of the network. Put another way, scalability refers to the amount of work that is required to increase the size or coverage area of an existing network. Scalability and ease of deployment are similar concepts, since it’s likely that a “new” network (a deployment) will have to integrate with an existing network (a scaling). The major difference between deployment and scaling is that deployment requires some amount of pre-installation planning, while scaling is more an extension of an existing plan.
Extensibility means that existing areas of connectivity can be joined together without requiring significant reconfiguration of either area. Extensibility comes into play when, for example, two companies partner with each other and it is necessary to provide connectivity between their existing networks. Extensibility is the quality that allows the two networks to seamlessly integrate with each other.
Central management means that interconnect devices like access points, routers, and switches across the network can be configured and monitored from a single location. Manual configuration on a per-device level is minimized. A global authority coordinates the management and configuration of the entire network. Even in cases where sub-sections of the network are locally administered (which is quite likely), these subdivisions conform to the mandates of the global authority. This ensures compatibility across the network.
None of these four characteristics exists in a vacuum; rather, each of them contributes to the others. For example, a network that is centrally-managed will also likely be easier to deploy and scale. Devices that support auto-configuration will be easier to deploy and scale, as well as easier to manage. Additionally, although we have been referring to “structured” versus “un-structured” networks, structured networking should be seen as a spectrum, rather than as an either-or state. A network should not be considered to be either “structured” or “un-structured.” Rather, we should evaluate the ways in which a network is more or less structured.
Pros and Cons Of Structured Networks
We have already described some of the benefits of more structured networks: they are simpler to design, deploy, and expand when compared to less structured networks. This simplicity translates directly to cost savings, but structured networks have additional benefits. The same qualities that make a structured network easier to design and install also make it less prone to breakdown. If the equipment is smart enough that it can automatically configure itself during installation, it is also likely to be smart enough to automatically re-configure itself in response to breakdowns. The centralized management characteristic of a structured network means that if problems do occur, the administrator will have an overview perspective on the entire network, making it easier and faster to pinpoint problem locations.
Structured networks do have some drawbacks. A structured network is likely to be somewhat less efficient than a less structured one. Because a structured network is designed using general guidelines, and because general guidelines tend to err on the side of safety, it’s likely that the amount of equipment that will be purchased will be somewhat more than the absolute minimum required to get the job done. In addition, building a structured network takes a conscious effort on the part of the administrator. In the long run, a structured network requires less planning than a less structured one because a single “cookie-cutter” guideline can be developed and then rolled out to all locations. But in the short term, the development of that “cookie-cutter” guideline can require more work up-front than simply installing the equipment and crossing your fingers (as is too-often done).
In general, we can conclude that structured design is most appropriate when the network will span many different sites, and the “cookie-cutter” guidelines can pay for their development time. If just a single site is to be deployed, it may be more efficient to create a less-structured design that is tailored to that site.
This article will be continued in next month’s newsletter. Stay Tuned!
Structured networking is a relatively new concept in wireless networks, but Ethernet networks have been designed using the principles of structured networking for years. The concept of "structured wiring" grew out of the original methods by which Ethernet stations were interconnected. Therefore, to explore the concept of structured networking, let’s start with an examination of two types of Ethernet—coaxial Ethernet (now essentially obsolete) which was essentially "unstructured" and twister-pair Ethernet, which formed the basis for developing a "structure" to the architecture of a network. Based on the discussion of Ethernet wired architecture we’ll see how these concepts can be applied to a wireless network.
The original Ethernet network, 10Base5, used coaxial cable as its medium. A single cable ran from the back of one station, to the back of the next, to the back of the next, and so on in daisy-chain fashion, as shown in the diagram below. This meant that the network designer had to carefully plan the most efficient path through the building that would allow a single cable to reach every machine while remaining within the maximum cable length. This made deployment difficult but, in the 1970's, there just weren't that many devices that needed to be interconnected!
A single coaxial cable connects all stations in a 10Base5 network. The yellow line represents the coaxial cable.
“Vampire taps” connect computers to the cable by piercing the outer sheath of the cable.
10Base5’s coaxial cable made it difficult to add or remove stations from the network. In a 10Base5 network, stations connect to the cable with a “vampire tap”. The vampire tap clamped down around the cable, permitting a pointy metal spike to cut through the shielding on the cable and touch the metal conductor in the middle of the cable. When stations were removed from the network, they left holes in the cable. A limited number of stations could be connected to a single cable. Once that limit was reached, a new cable had to be run to accommodate additional stations. These factors decreased the scalability and extensibility of 10Base5.
Adding a new station to a 10Base5 network would have required potentially inconvenient re-routing of cable.
10Base-5 Ethernet networks did not contain much, if any centralized management. The interconnect devices that made up these networks were generally configured via a serial cable that plugged into the back of the device or by TELNET-ing directly into the device. Each device had to be configured individually, and the devices typically did not support remote monitoring or statistics reporting. In those days, the best method of identifying the location of a problem was to literally split the network in half by disconnecting a cable and then see which side of the break still had the problem. Then that half of the network was divided in half again. This process, known as the “binary search” or “divide and conquer” method, was repeated until the source of the problem was isolated. Obviously, this was labor-intensive and had the additional disadvantage of interrupting network access until the problem was solved.
10Base5’s coaxial cable represents an extreme of un-structured networking. Running a single coaxial cable from station to station does not provide for the easy deployment and scalability that are the hallmarks of structured networking.
The next step in the evolution of wiring was the 10Base2 standard. It was still coaxial, but now the cable was thinner (more like coaxial television cable - and sometimes confused with it, to the chagrin of the installer!) Interconnection was done in the same daisy-chain fashion, but some implementations used 10Base2 hubs which allowed the coaxial cables to be run outwards in a star-wired configuration. The star-wired configuration was maintained as twisted-pair Ethernet entered, and then took over the market during the late 1980's and early 1990's.
Today’s Ethernet networks use switches with twisted pair cable as their medium. Switches are centrally located and a single cable runs from the station back to the switch. This usually ends up requiring a lot more feet of cable than a 10Base5 network would have, but twisted pair cable is cheap, and the ability to easily add new stations (by simply running another wire and plugging it into the switch) more than offsets the cost in extra cable. Plus, "10Base-T" (the original term for twisted-pair Ethernet) allows maximum separation of up to 656 feet (100 meters from hub to device, hence 2 devices separated by 200 meters which equals 656.17 feet)
A switched Ethernet network has a single cable running from the switch to each station, providing easier expansion and scaling than 10Base5.
It is easy to deploy a switched network. The only requirement is that each run of twisted pair cable is no longer than the allowable 100 meter maximum distance. Once you reach that distance, you simply plug in another switch and start another run of cable. This process can be continued essentially indefinitely (unlike the 10Base2 "hubs" which had a limitation known as the "5,4,3 Rule" - No more than 5 repeaters in line, interconnecting no more than 4 network segments, on which only 3 segments could contain end-user devices). Certainly, the performance of a switched network can be improved through careful design, but switched Ethernet networks can meet structured networking’s requirement of rapid, simple deployment without excessive planning.
Switched networks also have the structured characteristics of scalability and extensibility. If you run out of switch ports, you can simply add another switch and plug it into the first, easily extending the network. Two switched networks can be joined together simply by running a cable from one to the other. In most cases, no additional configuration is required in the switches.
Expanding a switched network is as simple as connecting two switches together.
Finally, switched networks are more likely to have the structured characteristic of centralized management, since enterprise-grade switches nearly always have a management interface that allows remote configuration and monitoring. Switches and routers also usually contain software that allows the device to automatically discover the network topology and create a basically-functional configuration upon startup. Often, these devices support protocols that allow an administrator to create a single, “master” configuration file and then push that configuration out to all devices at once.
A switched, twisted-pair Ethernet network is more conducive to structured design than one using coaxial cable because it is easy to deploy, scale, and extend, and because it is likely to support centralized management and monitoring.
The More Things Change, the More They Stay the Same...
Many of today’s wireless networks are, metaphorically speaking, similar to the 10Base5 Ethernet networks of yesterday. You have to manually manage the RF frequency space, making sure that access points’ coverage areas overlap enough that stations can roam between them, but not so much that they interfere with another transmitter on the same channel. Adding a new access point to the network is tedious because you have to configure all of its parameters individually. Managing the network requires monitoring each AP individually; there are no central points for management and security. But wireless networks are, in a way, the most conducive to structured design. After all, what could be easier to deploy and expand than a network that doesn’t have any wires to run?! What qualities would a structured wireless network have?
First, a structured wireless network should give the network designer the ability to easily predict how many access points and other devices will be needed to cover a certain area. Guidelines such as the number of users that a single AP can support, the aggregate throughput of a single AP, or the number of VoIP calls that a single AP can support assist in this assessment. A network that was installed based on these guidelines might not be quite as efficient as one based on a more rigorous site survey and un-structured design approach, but the savings in time and convenience more than make up for it. By way of analogy, switched Ethernet networks use a lot more wire than coaxial Ethernet, but the convenience of switched Ethernet means that the tradeoff is worth it.
In this month’s “Essential Wi-Fi,” we wrote:
In the long run, a structured network requires less planning than a less structured one because a single “cookie-cutter” guideline can be developed and then rolled out to all locations. But in the short term, the development of that “cookie-cutter” guideline can require more work up-front than simply installing the equipment and crossing your fingers (as is too-often done).
This statement requires re-examining from the perspective of a wireless network design. A wireless network design typically requires much more up-front work than a wired design. Not only do you have to do all of the capacity planning and needs assessment of a wired network, but you also have to figure out where to put the access points in order to provide coverage to the users. That process can easily take as much time and effort as several wired designs! In a wired network, you might get away with a perfunctory design and then “installing the equipment and crossing your fingers.” But in a wireless network, that approach will almost certainly lead to disaster. This means that even a small wireless network is likely to benefit from a structured design, since the effort of creating the “cookie-cutter” guidelines that can be applied to many locations is probably about the same as the effort of creating a tailored plan for a single location.
Second, the entire wireless network should be able to be monitored and configured from one location. You shouldn’t have to configure each access point individually—rather, the access points should auto-configure as much as possible. This is analogous to being able to plug two Ethernet switches together without having to specifically configure either one. Ideally, issues such as transmit power and channel selection are automatically adjusted by the access points in order to provide maximum coverage with minimum interference. Parameters that can’t be auto-configured should be able to be set in a central console that is accessible both locally and remotely. You should be able to monitor statistics for all access points from the same console.
Third, extending the wireless network should be as seamless as possible. Ideally, a single master configuration could be created that contained parameters that are common to all access points, and as many other settings as possible could be automatically determined by the AP. This would mean that new APs could be deployed quickly and with minimal manual configuration. But some manual configuration is inevitable in a secure network. A secure network requires that the station joining the network have some secret or unique piece of information. When the station wishes to join the network, it demonstrates that it has the secret or unique piece of information and is allowed to join. Unauthorized stations don’t know the secret and are kept out.
Security theory describes three types of “secret” that can be used in authentication: something you know, something you have, or something you are. An example of something you know would be a password or passphrase; an example of something you have would be a house key or an identification badge; an example of something you are would be your fingerprints or retinal pattern. Most wireless networks rely on “something you know”—specifically, an encryption key—for authentication. This presents a particular challenge to the goals of a structured network, since the key must be manually distributed and configured. This means that the goal of totally automatic extension of the wireless network is at odds with the goal of a secure network. Clearly, the goal of a secure network cannot be compromised, so the goal of totally automatic extension of the wireless network must be sacrificed.
This article will be continued in next month’s newsletter. Stay Tuned!
Ask the Expert
Less Antenna Gain = More Signal Strength?
In setting up a point-to-point link, I’ve found that I get higher signal strength with 16 dBi Yagi antennas than I do with 24 dBi dish antennas. What gives? Shouldn’t the higher-gain antennas give higher signal strength?
16 dBi Yagi Antenna (Roughly 20" Long)
19 dBi Parabolic Grid (Roughly 14" Square)
Higher gain antennas have narrower beamwidths than lower gain antennas. Their coverage pattern is focused in a smaller area, which means that they must be aligned more precisely in order to maximize signal strength. Lower gain antennas, on the other hand, are more tolerant of small mis-alignments.
The most likely scenario is that your antennas are not perfectly aligned. The Yagis have a wide enough beamwidth that the misalignment does not produce a noticeable drop in signal strength, but the misalignment is large enough to affect the dish antennas, with their narrower beamwidth. Although the antennas may appear to be perfectly aligned, RF line-of-sight is not always the same as visual line-of-sight. The antenna alignment that produces the maximum signal strength may not be the one that has the two antennas visually pointing directly at each other. Additionally, because of differences in antenna design, the optimum alignment for the Yagis may not be the same as the optimum alignment for the dish antennas, so switching from one antenna to another will require re-alignment.
Some readers might speculate that the problem is related to the Fresnel Zone. It is important to realize that the size of the Fresnel Zone is based only on the frequency of the signal and the distance between the antennas. The gain of the antenna does not affect the Fresnel Zone at all, so the link would have the same size Fresnel Zone regardless of whether the Yagi or the dish antennas were used. Therefore, the problem cannot be related to the Fresnel Zone.
The Connect802 team had a very successful showing at the BICSI Las Vegas show and educational forum. Joe Bardwell spoke at the event, discussing the current state of Wi-Fi, WiMAX, and CDMA/GSM cellular telephony in the marketplace. You can view the tradeshow video demonstration and obtain additional information about the RF CAD modeling and simulation services showcased in Las Vegas at www.connect802.com/info
More Unlicensed Frequencies by 2009?
Last month, the FCC set a road map for making additional frequencies available for unlicensed use by 2009. The new frequencies are currently allocated for licensed use by television broadcasters. Many of the frequencies are currently unused, and more will become available as broadcasters switch to digital television signals. Many of these frequencies are below 900 MHz, which makes them particularly desirable, as they are more able to penetrate walls and other obstacles than the 2.4 GHz and 5 GHz signals that are currently used by 802.11.
IEEE Broadcast Symposium Presentations on 802.22
In Washington, DC, presentations made at the IEEE Broadcast Symposium dealt with the proposed 802.22 standards for operation of unlicensed devices in the television broadcast bands; proposals that would create a Wi-Fi-like capability with a true range of possibly 30 miles, and the ability to easily penetrate buildings and structures. The standards, collectively know as Wireless Regional Area Networks (WRAN), are years from completion but target the 2009 anticipated changes in the way the FCC licenses the UHF television station frequencies. Engineers from the television broadcast industry expressed concerns over potential interference from unlicensed devices, and their concerns were met with discussion and various recommendations. At Connect802, we believe that standards like 802.22 will usher in the next major evolution in wireless communications. But, don't go looking for "pre-802.22" gear in your local computer store just yet. We're potentially looking at issues that will impact networks 10 years from now, at a time when today's Wi-Fi and WiMAX systems are approaching their end-of-life, or potentially their second fork-lift upgrade.
Web Searching: The Connect802 Web Presence
At Connect802 we're your PAGE ONE resource for wireless networking!
Connect802 has the experience, expertise, and resources to help you with your wireless network system. Use your favorite search engine and see what Connect802 is doing. Each month we give you some suggested search terms for you to explore. Here's this month's list. As you look down the search engine results you'll find Connect802 either at the top, or on the first page (true for Google and Excite, unknown for the rest).