The measure of reliability is the sum of the operating hours a population of a part operates divided by the number of failures. The reliability prediction can be determined by this example. Your manufacture 10,000 GLC-LX-SM-RGD-CDW transceivers. Each operates for 8760 hours (1 year) and 3 fail. Mean Time Before Failure = 10,000*8760 /3 = 87.6 million/3 = 29.2 million hours. So if Cisco knows how many GLC-LX-SM-RGD-CDW transceivers it sold at roughly what year the sold them and they know how many failed before shipping and how many were repaired as returns, an estimate can be calculated. What may be missing from my calculation is how many units were sold to replace failed units. Any sales volume by year would be helpful for this calculation.
