by Charles Brabham N5PVL
The suggestions here are taken from the very successful European FlexNet network. They have a great network, and here's why:
DAMA access, as mentioned in the various "General Information" articles here, eliminates Hidden Transmitter Syndrome (HTS). Eliminating HTS also eliminates the disadvanges of using low power packet stations. In fact, using the lowest power needed to connect to the node is preferable in an HTS-free environment such as DAMA provides.
Instead of having to use a high, omni-directional antenna and high power, you can use an H/T or low powered kit radio along with a yagi to access the nearest node. Things go faster since almost all packet collisions are eliminated, and it is easier to get on the air since it is no longer necessary for each station in the LAN to be able to hear each other. In the US, we have used the "everybody has to be able to hear everybody else" system for quite a while, and it has proven to have some serious limitations. Time to move on to something better.
If you start off using DAMA in the first place and do not provide old-style, non-DAMA access, the packet users in your area will adjust to this new system much more quickly. Starting off with DAMA will tend to encourage use of FlexNet by the most members of the LAN, which, in itself, will provide better overall performance.
Note: Owners of Kantronics equipment (which has DAMA capability) can still make DAMA connects even though they are not running FlexNet. (No FlexNet "6-Pack" driver for Kantronics equipment has been written yet, due to Kantronics' proprietary nature.) So even though owners of Kantronics equipment cannot run FlexNet, they can still participate in FlexNet's DAMA environment and there is no need for them to feel left out.
If you compromise and start off with old-style, non-DAMA access, you will find it very difficult to go over to DAMA later on. It's easiest and best to start off on the right foot in the first place. Use DAMA.
In general, it is desirable for node-node links to be at least eight times faster than the user access speed. For example: If user access is 1.2kb, then node-node (backbone) links should run at at least 9.6kb.
One way to speed up backbone links is to run higher speed, and another is to run full-duplex for these links instead of the half-duplex links commonly used for user access. Full duplex backbone links are very similar to two voice repeaters talking to each other's input freqs, in practice. Since the data flow is two-way and simultaneous, it ends up being significantly faster than a regular half-duplex link.
One example of the speed advantage with full duplex is that the data stream does not have to stop periodically so that ACKs can be recieved from the other station. Instead of going in fits and starts, the data flows much more smoothly and continuously. Using fullduplex does not eliminate packet link overhead, but it does greatly reduce it's effect by shunting most of that overhead off onto another frequency, where it will not interrupt the data stream.
European FlexNet nodes typically provide 9.6kb half-dup access, while utilizing 19.2kb fulldup backbone links. This is something to shoot for, as reliable connections of over 600 miles distance are common on the European FlexNet network, going through dozens of nodes. The faster your backbone links are in relation to your user access, the better overall performance you can expect from your network, especially as it grows.
If your only way to reach from A to B is through a long string of nodes, then an outage or problem at any one of those nodes will bring down that link. If there are other nodes that parallel that route though, they can be used to "route around" nodes that have problems and by so doing, eliminate or reduce the effect of service outages.
Use of redundant, parallel links in old-style packet networks in the US are uncommon, because they require either the node sysop to manually set up new routing when needed (no easy task!), or users to figure out the new routing for themselves... FlexNet, however, has advanced automatic routing built-in that transparently, automatically routes each connection by the fastest available way.
If a FlexNet node goes down, and other nodes capable of "routing around" that dead spot in the network exist, FlexNet "automagically" goes over to using the alternate route, transparently and without human intervention. When the dead node comes back online, the network begins using it again within five minutes.
Actually, this "automagic" routing is more functional than my description above implies... Since FlexNet knows the turn-around time to each node in the FlexNet LAN in so many milliseconds, it also "routes around" SLOW links automagically, not just dead ones. If propagation conditions, congestion, or some nagging problem temporarily slows down a node's backbone links and a faster route exists, FlexNet always takes the faster route. This gives congested links a bit of relief, speeding up everything considerably by putting part of the load on alternate routes.
For these reasons, FlexNet networks can take especial advantage of "redundant", parallel links, and the setup of these secondary paths should be seriously considered by FlexNet node sysops. In metropolitan areas, a series of FlexNet nodes will serve much better than a single node set up in a high place. This setup works in much the same way that "voting" voice repeaters operate, providing better access and more "bulletproof" routing. On long-haul routes, instead of using a long string of high nodes, a greater number of less capable nodes comprising a "web" can actually provide better service. Long-haul routes can benefit greatly by placing redundant nodes off to the side that can take up the slack during times of congestion and provide instant backup during node outages.
Use of these methods will give you a significantly faster and more capable network. FlexNet has many unique abilities and by using DAMA along with high-speed, redundant links, you can get the most out of your FlexNet network.
Any of these methods by themselves will serve to speed things up, but taken as a package they will provide truly amazing performance; much better than US packet ops are used to seeing in an amateur packet radio network.