Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Posting-Version: Notesfiles; site hpfcms.UUCP Path: utzoo!watmath!clyde!burl!ulysses!gamma!epsilon!zeta!sabre!petrus!bellcore!decvax!decwrl!pyramid!pesnta!hplabs!hpfcdc!hpfcla!baker From: baker@hpfcla.UUCP Newsgroups: net.arch Subject: Re: Re: IOCALL results and problems Message-ID: <14600002@hpfcms.UUCP> Date: Fri, 3-Jan-86 16:36:00 EST Article-I.D.: hpfcms.14600002 Posted: Fri Jan 3 16:36:00 1986 Date-Received: Mon, 13-Jan-86 08:00:41 EST References: <354@ncr-sd.UUCP> Organization: 03 Jan 86 14:36:00 MST Lines: 30 Here at HP we use a system to test the multiuser performance of our mini/micro systems using a scheme much like that described in previous notes. The system is called termial emulation and performance evaluation (TEPE). One or more HP3000's are used to simulate users on the system under test (SUT). Scripts were developed to simulate the behavior of users in several different environments (program development, word processing, computationally intensive, etc.). TEPE generates a random inter-character typing delay and a random "think time" between commands given time ranges for each interval. This random delay coupled with a staggered start time for each "user" on the SUT, prevents corrupt timings due to lock-stepped script execution. The script is iterated until the specified time interval has elapsed. Measurements taken at the start of the test are discarded because not all "users" have instantiated a session and the load has not stablized. Tools for collecting, summarizing and plotting the data have been developed to allow performance characterization up to 64 users. Obviously, this is not a standard test that many vendors publish as performance data. So, if HP is interested in a true multiuser comparision to a competitors system, a machine must be obtained for the duration of the test. TEPE is used largely used within HP for comparing internal products and for tuning. Jim Baker hpfcla!baker Fort Collins System Division