On "burn-in".
The underlying principle is that for every system of any complexity, most certainly electronic systems and computers, the likelihood or probability of failure over a given time interval is typically highest at the beginning of the system's life.
This probability of failure then drops rather quickly with time to reach a (let's hope) a low level holding over a long period of time which corresponds to the "useful lifetime" of the system. After this lengthy low level, the probability of failure (over a fixed time interval) will climb again signaling the end of the system's useful life.
To put figures on this phenomenon is what would be required to answer the question of burn-in requirements. Certainly manufacturers gather a lot of statistical data to quantify this reliability process. In fact, this ties in with their warranty and one may assume that they would not let a complex system as computers out of their factory without making sure that the system is beyond its "first hours of life" where failure would be most likely and cost them money for servicing. It would be safe to assume that for typical use no burn-in is necessary as the manufacturer has probably not taken any risks on warranty.
If the user requires a very high level of reliability, it may be necessary to observe the system's behavior over a sufficient time period. In this case meaningful statistical failure data needs to be available in order to say anything meaningful about such requirement.
The underlying principle is that for every system of any complexity, most certainly electronic systems and computers, the likelihood or probability of failure over a given time interval is typically highest at the beginning of the system's life.
This probability of failure then drops rather quickly with time to reach a (let's hope) a low level holding over a long period of time which corresponds to the "useful lifetime" of the system. After this lengthy low level, the probability of failure (over a fixed time interval) will climb again signaling the end of the system's useful life.
To put figures on this phenomenon is what would be required to answer the question of burn-in requirements. Certainly manufacturers gather a lot of statistical data to quantify this reliability process. In fact, this ties in with their warranty and one may assume that they would not let a complex system as computers out of their factory without making sure that the system is beyond its "first hours of life" where failure would be most likely and cost them money for servicing. It would be safe to assume that for typical use no burn-in is necessary as the manufacturer has probably not taken any risks on warranty.
If the user requires a very high level of reliability, it may be necessary to observe the system's behavior over a sufficient time period. In this case meaningful statistical failure data needs to be available in order to say anything meaningful about such requirement.