There are four ways to measure and describe reliability, each subtly different. Reliability, measured as R(t), is a probability, the odds that a supply operating under a specified condition will still be working after a given period of time. Conversely, failure rate is the percentage of units that fail in a given unit of time. It almost always follows a so-called "bathtub" curve. Two other useful measures are MTBF and MTTF, defined as 1/λ. MTBF is useful for equipment that will be repaired and then returned to service, but despite the commonplace assumption, it does not guarantee a minimum time between failures, only a mean. MTTF is technically more correct mathematically, but the two terms are equivalent and MTBF is the more commonly used. Additionally, service life is the amount of time that the supply needs to operate in its intended application. A long service life does not necessarily correlate to a long MTBF, and some applications need a large MTBF but only a short service life. Picture a 3-second experiment taking place on a space station. The cost of repair is prohibitively expensive so a high MTBF is essential, though the required service life is exceptionally short.