Network Computingwww.networkcomputing.com

Some backup connections may be operated at a minimal cost until data actually flows. Once the routers begin using the alternate path, though, carriers may charge you per cell or per kilobyte. If you're being charged on a per-kilobyte basis, you want to minimize the amount of time data is running on the backup network. And because establishing PVC (permanent virtual circuit) connections sometimes can take several minutes, there are trade-offs to consider. If your network experiences a brief hiccup, you don't want the router to become overzealous and switch immediately, only to have the main network back online just as the backup becomes available.

If you use your secondary connection to handle live traffic, this is less of an issue: The PVCs are in place and the switch-over is immediate, limiting downtime. Traffic or ports can be weighted to determine when a different path should be tried based upon preset time-out conditions. Routers also can be set not to switch to a secondary port unless an RDI (remote defect indicator) or loss of LMI (local management interface) occurs on the primary port.

If your carrier's network should fail, having backup connections to the network won't do you any good because the links all go to the same place. A different backbone must be used. In mission-critical situations, where a backup is absolutely necessary, the only solution is to have secondary connections to another carrier's network. These connections would always be up and available on a moment's notice. This solution gives you not only a backup local loop but also a backup backbone, providing the best all-around solution. Expense on the second carrier can be kept to a minimum by purchasing 0-CIR (committed information rate) frame relay or ABR (available bit rate) ATM PVCs.

At What Cost Convergence?

Of course, providing a backup connection for your local loop flies in the face of the hoopla surrounding convergence. The goal of convergence, after all, is to transport all your data and voice communications on one connection, simplifying the wiring closet and the links coming into your office. Vendors and carriers have been touting convergence for years, but that doesn't mean you should put all your data on one pipe. Convergence can become a liability for some enterprise customers: Voice and data convergence could cause a single point of failure that could devastate business.

Local-loop backup has been around ever since vendors added BRI (Basic Rate Interface) ports to their routers. But back then most frame relay connections were less than 256 Kbps in bandwidth. Backing up that bandwidth temporarily with 128 Kbps or even 64 Kbps wasn't much of a hardship. Even the ability to dial into a carrier frame relay network with an analog modem was available.

Although this method was used primarily as a type of VPN (virtual private network) by mobile users, it could have been extended as a backup for the office. But expecting an analog or ISDN line to carry traffic normally destined for a connection of 3 Mbps or greater is wishful thinking, and trying to make it happen can cause major network slowdowns. Just how much bandwidth needs to be provided for depends on the urgency or importance of the data. Your backup should be able to handle 75 percent of your average essential bandwidth. If you have 2 Mbps of essential data traffic, you should have a full DS-1 ready as a backup connection.

Providing a secondary connection for a high-speed local loop becomes a little more difficult than backing up a DS-1, and any such solution is going to be more expensive than keeping a standby ISDN or analog line.

The types of network outages that should be considered are soft and hard loop outages, and outages in the network core. Soft outages include excessive line noise or problems with the local switch that cause the frame relay LMI to go down. Hard outages can be classified as switch/router hardware going four paws to the sky or the local ditch digger not looking while excavating.