O'Reilly - rulez!
Вот чем, мне нравятся книжки издательства O'Reilly, так это тем, что не смотря на занудность предмета, они написаны таким простым языком что читаешь как художественную литературу. Вот например "JUNOS Enterprise Switching", глава High Availability:
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. 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, just as you think you have a total grasp on HA, another wrench is thrown into the puzzle that may affect your measurements. The lesson learned: be aware of the factors that are measured when one vendor claims 99.99999% HA while another claims 99.99%. Is it really an apples-to-apples comparison?
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. 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, just as you think you have a total grasp on HA, another wrench is thrown into the puzzle that may affect your measurements. The lesson learned: be aware of the factors that are measured when one vendor claims 99.99999% HA while another claims 99.99%. Is it really an apples-to-apples comparison?