Only a simulation

ITworld.com –

Simulators are all the rage these days. There are flight simulators, driving simulators, city simulators -- and you and I have network simulators.

Although the first items I listed are diversions for most, network simulators are serious business and can come with a serious price tag -- as much as $30,000. That's certainly not pocket change, but it can be cheap compared to the cost of building a prototype network using real hardware or implementing a new design on a production network.

Network simulation packages let you model enterprise network topologies and enterprise applications. You can use them to estimate a network's performance and capacity. They allow you to create what-if scenarios, which can be useful in provisioning bandwidth and adding new network devices such as switches and routers.

If a simulator can accurately model a production network, it solves a quandary confronting network architects and operators. Testing new equipment or designs on a live production network is never advisable. On some networks, mission-critical applications run 24 hours a day, seven days a week, and you don't want to be testing when someone is depending on the network. Testing can result in short-term or (perish the thought) long-term outages.

As a result, new designs are often relegated to a test network that tries, with varying degrees of success, to replicate the production network. A scenario may work just fine on the test network, but may collapse when moved to the production network.

Network simulators offer a way around this dilemma. Simulators use mathematical models with a GUI front end. Most contain templates that model network devices, including the interfaces you have on the real network and the infrastructure that connects them. If the model is accurate, you can create networks that are good simulations of your actual network. New designs and concepts can be safely modeled using network simulators without endangering the actual network.

Many network simulators will automatically discover and import the devices and topology of the production network, though not necessarily the traffic level. You can specify traffic levels to test on the various simulated segments.

However, importing a snapshot view of the network may not always be a practical approach. One reason is that networks are never static. From the time a model is imported to the time a scenario is ready for testing, the network may have undergone major changes in infrastructure or traffic flow. But an even more important reason to avoid the import function is that it's not always as accurate as the vendors allege.

Rather than import the network, which can take plenty of time itself, you can manually construct it from templates. This allows you to create the network as you know it at a given moment. Both approaches give static pictures, but the manually created network may be more accurate than the automatically imported network, particularly if you begin running stimulations immediately after you create the network from the templates.

Another downside of network simulators is that they take a long time to learn to operate properly, and each simulation run can take a long time -- often several hours. Time can be tough to find if your network staff is overtasked, as is often the case.

When you combine the resource requirements with the high cost of these packages, they can be a tough sell to management as well.

Because of all these factors, simulators make the most sense for those designing and operating large enterprise networks or WAN-oriented services.

If you're interested in investigating these products further, here's a list of candidates to get you started:

Insider: How the basic tech behind the Internet works
Join the discussion
Be the first to comment on this article. Our Commenting Policies