12.1 Traffic-Engineering Problems

The Internet's level of complexity has rendered simple IGP-based methods of controlling traffic flow insufficient. Figures 12-1 and 12-2 demonstrate why it is not always best to use metrics to control traffic flow. In Figure 12-1, traffic from Chicago destined for Tampa and Miami is crossing the link from Chicago to Atlanta. This path is chosen because it has a lower metric than the path through Washington D.C. In this situation the link from Chicago to Atlanta could become overutilized, while the path through Washington D.C. remains underutilized .

Figure 12-1. The Underutilized Path Through Washington D.C

Figure 12-2. The Underutilized Link from Chicago to Atlanta

In Figure 12-2, an attempt is made to force use of the path through Washington D.C. The metric for the link directly between Chicago and Atlanta is raised to 5. This forces the use of the path through Washington D.C., which has a combined metric of 2. With these settings, traffic destined for both Tampa and Miami will follow the path through Washington D.C., which will become overutilized, while the link directly from Chicago to Atlanta is underutilized. This change did not resolve the issue; it simply moved it.

One way to resolve the dilemma shown in Figures 12-1 and 12-2 would be to have traffic destined for Tampa use the direct link from Chicago to Atlanta and to have traffic destined for Miami follow the path through Washington D.C., as shown in Figure 12-3.

Figure 12-3. Both Routes Utilized

Figures 12-1 and 12-2 demonstrate the lack-of-control issues faced by network engineers . Figure 12-3 represents the essence of traffic engineering as a means to resolve the control issue. In Figure 12-3, neither link is underutilized, and traffic is routed by the Chicago router, based on the final destination of the traffic (Tampa or Miami). The next section focuses on traffic-engineering solutions.