网络工程规划设计需求分析.PPT

* * * * * * * * * * * Resiliency means how much stress a network can handle and how quickly the network can rebound from problems, including security breaches, natural and unnatural disasters, human error, and catastrophic software or hardware failures. Some experts, including Howard Berkowitz, have a mild dislike of the word “resiliency” as it sounds too much like a stretched rubber band or a trampoline. As Berkowitz says in his excellent book, WAN Survival Guide (Wiley 2001), “I avoid designing networks that stretch too far, bounce up and down, or oscillate between normal and backup states.” So he likes “fault tolerance,” but he points out that it does not mean “immune to any conceivable threat.” Berkowitz states that, “A sufficient quantity of explosives can overcome the tolerance of any network.” :-) * * * * * * 99.70% availability sounds pretty good, but it could mean that the network is down for 0.18 minutes every hour. This is 11 seconds. If those 11 seconds were spread out over the hour, nobody would notice possibly. But if there were some bug, for example, that caused the network to fail for 11 seconds every hour on the hour, people would notice. Users these days are very impatient. Notice that 99.70% availability also could mean one catastrophic problem caused the network to be down for 1577 minutes all at once. That’s 26 hours. If it were on a Saturday and the network was never down for the rest of the year, that might actually be OK. So, you have to consider time frames with percent availability numbers. Consider the holy grail: 99.999% availability. That’s 5 minutes downtime per year! Be sure to explain to the customer that scheduled maintenance and upgrades don’t count! Either that or plan for a network with triple redundancy (that could be extremely expensive to implement and operate). * In the event of failure of the primary router, the secondary becomes the primary and still has a backup. Fix the previous primary and have it become the tertiary. T