15.5 Packet Data Network Recovery

---

Frame relay is a popular data service for interconnecting LANs at multiple corporate locations. The data are sent over virtual circuits in the service provider’s network, which are priced on the basis of an assigned minimum increment of bandwidth (i.e., CIR) rather than distance, making frame relay more economical than private lines, especially for large mesh topologies.

One way to protect frame relay virtual circuits from loss due to a catastrophic failure within the carrier’s network is to subscribe to a dual frame relay service in which data traffic is load-balanced across two frame relay networks to provide businesses with uninterrupted service. The secondary network mirrors a company’s primary network, providing separate local-access lines, switch ports, and PVCs. Customers can seek either full or partial redundancy to their primary network without needing to diversify service across multiple carriers for optimal network reliability and availability.

Another protection method guards against the loss of a crucial port, which can be disastrous to a hub-and-spoke network configuration where multiple remote sites (spokes) are tied into a host site (hub). To protect the network if the host port goes down, carriers build PVCs from one or more remote sites to an alternate host port. When these PVCs are activated upon request, traffic is redirected to the backup host port within minutes. When the primary port is reestablished, traffic resumes to the primary host location. There is no need for complex rerouting algorithms within the CPE. The carrier’s restoration plan is tested to ensure proper operation when needed. In addition, customers can request tests on activation plans without charge. Several carriers offer this type of protection service for their frame relay customers.

One of the attractive aspects of using frame relay is that it offers more built-in reliability than leased lines such as T1. Although it does not offer error correction—this and other overhead functions of X.25, a predecessor packet technology designed for noisy analog networks in the 1970s, have been stripped out—frame relay has a dynamic rerouting capability. If a trunk on the carrier’s network goes down, the PVCs can be set up to be automatically rerouted over different trunks. In contrast, when a leased line goes down, traffic often must be manually rerouted to another available line with spare bandwidth.

The features of some frame relay switches maximize this inherent reliability to the point of creating a fault-tolerant frame relay network. These features include the following:

• Fault-tolerant PVC: Provides automatic cutover to a designated set of backup ports on the frame relay switch, rerouting all affected frame relay circuits from a failed data center to a backup data center, for example.

• Fault-tolerant trunking: Permits the frame relay switch to automatically establish a backup trunk to reroute all circuits over an ISDN network in the event of failure.

• Access failure recovery: If a branch location loses access to the primary frame relay connection, an ISDN dial backup call is automatically placed to another frame relay switch.

Another type of packet data network relies on ATM technology. ATM uses very small fixed-length packets (also called “cells”) to achieve speeds in the gigabit-per-second range. At such speeds, traffic control is needed, especially since connections for each call may be routed through any number of intermediate switches. Each switch must be able to handle all pass-through traffic. Congestion problems occur when the input rate of traffic into a switch exceeds that switch’s available link capacity. Long-duration congestion is not necessarily a function of switch design; it just might be better handled with improved network design or increases in overall network capacity. Short-duration network congestion often can be alleviated simply by configuring the switch to buffer inbound data. ATM’s rate-based congestion control scheme provides end-to-end control using single-bit feedback. The switch monitors the network by way of these feedback bits and, when traffic is detected, adjusts the data rate up or down until the problem goes away.

High-speed ATM services over the WAN are implemented over highly reliable optical-fiber facilities, specifically SONET, which confers a high degree of reliability to ATM. Not only can SONET rings protect ATM traffic with its sub-50-ms service restoration capability, the embedded management channels of SONET can be used to detect real or impending performance problems. Consequently, ATM does not need much in the way of management functionality beyond a simple congestion control mechanism in the carrier’s network and QoS parameters implemented at the customer’s equipment and carried through the ATM network at initial configuration.

---