JUNOS Enterprise Switching

Chapter 11. High Availability

High Availability (HA) has been a hot topic ever since two devices were networked together over a piece of barbwire. What happens if the wire breaks? What is the backup plan, and how long will it take to solve a problem? That’s the essence of High Availability: ensuring some level of reliability and redundancy. One of the difficult elements of HA is measuring the success of and understanding all the components involved. Is HA measured on a time factor, a packet loss, or a delay? The answer is all of the above, so when 99% reliability is advertised, you must look at the factors that go into the calculation of that measurement. To look at a non-network example, imagine you own a bar in San Francisco that makes top-notch margaritas.^[3] You need to ensure that the drink tastes the same each time it is served. This requires the ingredients to be readily available and consistent. Some of the factors that could affect a margarita include:

• Type of tequila

• Number of fresh limes

• Amount of agave nectar

Many of these items can be quantitatively measured; we could say that if all the ingredients are measured and are available every day, 99% of the time the drink should taste the same. However, could any other factors affect the taste? The quality of the lime, for instance, or the quality of the agave nectar? Perhaps the type of food the person was eating at the same time could affect the taste. These variables could significantly lower that 99% value. The point is, ...