Converged networks typically require a higher availability metric than non-converged networks (there are exceptions, typically in the financial world). Five-nines is often referenced as the basic level of availability that traditional telephony networks used as their metric. If you do the math on 365 days X 86400 seconds/day, that winds up being 5.25 minutes per year of downtime.
Network availability can be measured any number of ways, some of which make it much easier to achieve five-nines availability. If I look at any outage that affects more than 10% of the customers and sum that over the year I get one number. But I can also say that since it affected only X% of the customers, I get to pro-rate it by that same percentage, which makes it much easier to achieve a higher availability value.
Whatever the methodology for calculating the availability metric, there are a number of fundamental things that you can do that will improve a network and its availability. I’m writing about what you can do in a series of blog posts at nojitter.com, starting with a blog “Want a Five-Nines Network?”
The first blog post talks about what you can do in the network itself. The foremost things are to run one OS version per hardware platform and to keep the hardware up to date. Yes, this can be expensive. Running a five-nines network does cost extra money. But if it makes more money as a result, it is a good investment.