Fault-Tolerant IP and MPLS Networks

10 Chapter 1: Understanding High Availability of IP and MPLS Networks

Network and Service Outages

A service is the set of tasks performed by the network upon a request from the user such as a

voice call, Internet access, e-mail, and so forth. A service outage is the users’ inability to request

a new service or to continue to use an existing service because the service is either no longer

available or it is impaired. As discussed previously, availability of a network strongly depends

on the frequency of service outages and the recovery time for each outage. A network outage is

the loss of network resources, including routers, switches, and transport facilities, because of

the following:

• Complete or partial failure of hardware and software components

• Power outages

• Scheduled maintenance such as software or hardware upgrades

• Operational errors such as configuration errors

• Acts of nature such as floods, tornadoes, and earthquakes

Planned and Unplanned Outages

Each network outage can be broadly categorized as either “unplanned” or “planned.” An

unplanned network outage occurs because of unforeseen failures of network elements. These

failures include faults internal to a router’s hardware/software components such as control-

plane software crashes, line cards, link transceivers, and the power supply or faults external to

the router such as fiber cuts, loss of power in a carrier facility, and so forth. A planned network

outage occurs when a network element such as router is taken out of service because of

scheduled events (for example, a software upgrade).

Main Causes of Network Outages

What are the main causes of network outages? As it turns out, several culprits contribute to

network downtime. According to a University of Michigan one-year reliability study of IP core

routers conducted in a regional IP service provider network, router interface downtime

averaged about 955 minutes per year, which translates to an interface availability of only

0.998.

3

As a reference point, a carrier-class router is expected to have a downtime of only 5.2

minutes per year. The same study indicated the following percentages of causes for total

network downtime:

• 23 percent for router failure (software/hardware faults, denial-of-service attack)

• 32 percent for link failures (fiber cuts, network congestion)

• 36 percent for router maintenance (software and hardware upgrade, configuration errors)

• The remaining 9 percent for other miscellaneous reasons