試験CTFL-PT トピック3 問題16 スレッド

A. If the quiz application drops more than 1% of quiz sessions, users will likely write bad reviews on app stores and other gaming sites Performance tests will verify that system can support at least 4 95 million users Planned tests, which include the simultaneous sessions test, will cost $350,000 Since only two quizzes are scheduled a day, performance tests on the production system during the "down' times should be representative of what users experience. There is a 10% chance that the production system will not restored in time for a quiz.

B. There is a less then10% chance of missing a scheduled quiz session due to performance test activities, which will likely cost the company $50,000 in lost sponsor revenue and brand damage To avoid this risk, we are scheduling Operations and Development representatives to be available during test execution We will require that the RAID 5+1 LUNs associated with the database server cluster be broken at 22 30 and the RAID 5+1 LUNs restored, ensuring the offline set mirrors back to the RAID 5 set that was under test, at 03:00

C. During the actual quiz, users logon gradually but play simultaneously. The load injection of quiz responses will not ramp up. therefore nearly 5 million quiz responses will need to be generated in a 4 second time frame with no opportunity to reach steady state.Test plans have been discussed with Operations Development and Data Center representatives and we believe the tests will tax the quiz application consistent with actual use.The JMeter monitoring system has been tuned however, with 5 million service requests over 4 seconds there is concern that the services queue measurements may be off by 25%

D. Since the client uses an asynchronous REST web services based sessions back to the QUZRAC services cluster, the loadprofile will consist of virtualusers running programmed load scripts based in C# that will randomly select the correct answer following a normal distribution thus modeling the typical success distribution among users The strict 4 second cap on user responses will translate into think times that are randomly selected between 1/2 second and just under 4 seconds using a uniform distribution