In ur test 90℅ of transactions met tat response times. Say ur running load test for 1hr n u having min avg max response time values but those are like some times server take more or less time to respond to user request. The 90th value is like how many times in tat 1hr server responded with same response value say like ur avg response time is 2sec and 90th value was 3sec so here server 90℅ of transactions got response in 3sec.
90 percentile is helpful when we run a long duration test. Suppose you ran a test with 100 iterations, the LR will arrange response time for those 100 iterations in arranging order and the value which comes in 90th position will be 90th percentile. 90 percentile means 90% of the transaction response time is less the value what you got at 90 position. We consider average response time when we have less number of transactions and 90 percentile when we have large number of transaction count.
Due to worst transactions average response time will be impacted &can't report the same,to measure the application transaction RT we have to consider 90% RT Ex: Log in transaction executed 10 times & reported below 1 sec for 9 times. 1 time it reported 100 sec, due to 1 occurance avg response times are very high. to overcome this situation we have to calculate 90% RT.
