Home > Focus > Full Story

Focus Wireless Deployment

Wired for Wireless

Although wireless is a great way to connect, the implementation issues can be daunting. A look at the key challenges for wireless deployments. by Graeme K. Le Roux

Lately, major PC vendors have been pushing "wireless network infrastructure" in the market place. The impetus for this push has been a mixture of new technology in the form of 802.11b adapters being available in quantity and the usual need for a "new" feature to market (read: hype).

Various low speed wireless systems have been around for the best part of ten years while several proprietary systems, mostly application-specific types, pre-date them by several years.

The major difference, from a design standpoint, between working with systems which incorporate wired components and exclusively wired systems, is the switching architecture. Unlike their wired brethrens, wireless networks are usually bridged rather than switched.

Wireless systems are often deployed because they allow users to roam about. If a user is working on a laptop in one place, then a wireless connection is logically similar to a simple 10Base-T Ethernet link, where the NIC in the laptop has a message authentication code (MAC) layer address, and the bridge has a port to which that NIC is connected. Also, the bridge maintains a table which tells it which port the NIC is connected to.

But what happens if the user picks up his laptop and walks to the other side of the building? The NIC still has the same MAC layer address, but it is not connected to the same bridge port, in most cases it is not even on the same bridge. Obviously there has to be some way of handing off the NIC from one bridge to another.

In an 802.11b environment, each bridge has a discrete network ID (also known as a Radio Group, and not to be confused with an IP Net ID) which it shares with all the NICs it is communicating with. A hand off from one bridge to another is based on the signal to noise ratio (SNR) as measured by the roaming NIC. The SNR is measured on the basis of beacon signals transmitted by each bridge in the LAN as SNAP frames.

These SNAP frames also contain information about the bridges' network IDs, and so on. When a roaming NIC needs to initiate a hand off, it first finds a new bridge then signals its current bridge requesting a hand off to the new bridge. The bridges then contact each other via their common backbone connection using SNAP frames and update their tables. Finally, the NIC changes its network ID to match that of the new bridge. Actually, the NIC's MAC layer ID has not changed, just the ID associated with the bridge it is communicating with.

Of course, the physical location of the NIC may change quite frequently and this can be a problem in an IP-based network-since IP implicitly assumes that IP addresses are geographically fixed. Specifically, an IP address is composed of a host ID and a net ID. Routes which are used to forward data are sets of instructions for getting from one net ID to another.

For example, if a user whose laptop is configured with an IP address of 10.0.1.2 mask 255.255.255.0 (that is, the host 2 on sub-net 10.0.1.0) moves to a wireless bridge which is connected to IP sub-net 10.0.2.0, it will not receive data.

To allow communication, we have to change the laptop's IP address when we move between access points which are attached to different IP networks. Consider a user with a laptop containing a wireless NIC in a multi-storey office block in which each floor is a different IP sub-net and each floor has four wireless bridges (referred to more commonly as access points). The user can happily roam anywhere within any single floor of the building with a given IP address, but as soon as they move to another floor they will not be able to communicate unless they change their IP address.

Since having users manually and routinely fiddle with their IP address is every network administrator's nightmare we have to have a way of assigning IP addresses automatically. This can be done with dynamic host configuration protocol (DHCP), but there are limitations.

DHCP leases IP addresses from a pool. With it, clients in the scenario described above have to request a new address after checking and finding that the one they have is invalid. Unfortunately, this sort of check only happens when a lease times out, or when the client's IP stack is initialized. Depending upon how the operating system on the user's laptop implements the IP stack, this may happen when the laptop comes out of a "suspend" or "sleep" mode and will certainly happen when it is rebooted. In practice, this means that users may have to suspend their laptops, or reboot whenever they change floors.

For laptop users, a reboot would be inconvenient. However, putting it to sleep may be viable. Besides, it is unlikely that people will be using their laptops in a lift or while walking up or down the stairs. There is one catch though: you can't assume that your wireless network user will be using a laptop-it might be a PDA or in future some sort of mobile VoIP phone. Suspension may not be an option in such cases. For that, you'll need mobile IP.

Mobile IP
Mobile IP (RFC 2002 and updates) is a proposed standard which assigns each mobile host to a "home agent"—basically a router which is on the IP network where the mobile host is notionally based.

Take the building example above. If the "home agent" is on the second floor, then routers on the other floors are the "foreign agents". The home agent keeps track of the mobile host's current location and in conjunction with the foreign agents, tunnel data to the mobile host. This implies an overhead on the routers and the wired network backbone, but it should not be too much of a problem since the wired system can always be provisioned beforehand to cope with it. Actually, until the full standard specifications of mobile IP are available (only pre-standard ones are around at present), it is still too early to speculate on what extra overheads mobile IP will create.

What are some good strategies to adopt when building wireless networks? A simple way is to employ switches and bridges at internal nodes, and routers at the edge of your system. However, if you have a multi-storey implementation, having routers on numerous floors can be expensive. A more cost effective solution would be to interconnect all the wireless bridges via a system of full-duplex 100Base-Tx Ethernet switches.

Note that while having several 802.11b bridges per floor creates overlapping signal areas, this does not introduce loop paths in the network because each 802.11b bridge has a different wireless network ID and thus are logically connected via the wired backbone and not a wireless link.

But wireless backbones can be created if desired. Wireless bridges with more than one port can be configured such that each port has a different network ID. By configuring one port on each of multiple ports with the same network ID, you can create point-to-point links that can be employed as a wireless backbone instead of using a wired system.

While this is low capacity compared to either UTP or fibre, wireless backbones compare favorably to short haul leased line services and eliminate the need for cable. This can be useful if you need to connect say, solar-powered information kiosks scattered in a large public area.

In fact, for a small system you need not even consider the use of bridges and wired back bones. A small ad-hoc wireless LAN can be created using 802.11b simply by bringing a number of wireless NIC equipped hosts within range of each other and setting each of them to have the same wireless network ID.

Hey, this means that you can hold the company board meeting in the local park. Or better still, in a half-decent pub.

Graeme K. Le Roux is the director of Morsedawn (Australia), a company which specializes in network design and consultancy and writes for Network Computing-Asian Edition. He can be reached at graemel@moresdawn.com.au. Send your feedback to editor@networkmagazineindia.com

Wireless pros and cons
Wireless technology is generally intended to be used in locations where it is too expensive or impossible to deploy a wired solution. An example is a warehouse environment where terminal devices are mounted on mobile platforms such as fork lifts. Other examples include exhibition spaces where the cost of re-wiring for each exhibition would be too expensive, and heritage buildings where drilling of holes for cables is not permitted.

Wireless technology in the form of 802.11b can be used as a cheap point-to-point link for campus environments, usually by adding an external high gain, uni-directional antenna.

The biggest downer to wireless is its lack of speed. A wireless network is necessarily half-duplex, which means that at best, an 802.11b system will provide performance equivalent to a wired half-duplex 10Base-T system. If you are running full-steam with wired networks, don't expect to move the same applications to wireless without experiencing stutter.

Wireless systems also necessarily complicate the design of large networks, and wireless network adapters are more expensive compared to wired ones (although cabling costs are reduced). Whether this reduction in cabling costs counters the added cost of the adapter depends on the specific situation in which you are thinking of deploying wireless.

802.11b NICs use direct sequence spread spectrum technology which minimizes interference problems. Most NICs also support some form of encryption. The aim is "wire equivalent" security. That means roughly as good as you could expect in the average office with a UTP Ethernet connection. If you don't encrypt traffic on your wired Ethernet, you don't need more than the security available in most 802.11b NICs. If you do, then you can use the same encryption as you use in your wired LAN over 802.11b.

IEEE 802.11b equipment basics
Most vendors build 802.11b adapters as PC cards which can be directly inserted into laptops or their access points (bridges). The same cards can also be inserted into adapters to provide ISA and PCI bus support.

Data rates in all wireless systems depend on the distance between the client and the access point and any obstacles in between. Typically, vendors use the 802.11b HR standard which delivers 11 Mbps over a maximum range of between 25 and 125 m with a standard antenna. Such cards can fall back to 5.5, 2 or 1 Mbps to interoperate with older systems or to cover longer distances—anything up to roughly 500 m with a standard antenna in open country.

Most vendors also either supply or provide for a number of alternative antennas, usually intended for attachment to an access point, which can allow for coverage over distances of between 1 and about 25 km.

Most access points support 10/100Base-T Ethernet via UTP. Some units are also available with a port for, or a built-in, WAN link. This usually ranges from a V.90 modem to most kinds of retail broadband connection (cable, xDSL, etc.)

Most access points are shipped with diagnostic software which will, in conjunction with a laptop and NIC, allow you to "qualify" a site-that is, check how many access points you will need and where you will have to put them. This is a basic and necessary part of the planning process. Installers worth their salt will qualify a site prior to giving a quote on supply and installation of equipment.

Some vendors also provide powered Ethernet options. This is simply a way of providing an access point with DC power via its UTP back bone connection. This makes it much easier to install an access point in an awkward place such as inside the false ceiling in the middle of a room far away from a wall socket.

Resources
For more on the use of 802.11b in LAN and networking environments, visit the following Web sites:
http://www.wirelessethernet.com
http://www.ietf.org
www.WirelessEthernet.com (see under "member companies")

 

>>Next