From: broehl@sunee.uwaterloo.ca (Bernie Roehl) Subject: Re: SCI: Networked VR Models Date: Tue, 25 May 1993 14:17:11 GMT Organization: University of Waterloo In article <1tk548INNrsk@news.u.washington.edu> Jesse Schell writes: >[...in reference to performance metrics for networked vr systems...] >Here is the list at present: >--frame rate This is independent of the network, and depends more on the capabilities of your graphics hardware. >--lag (the time from moving a sensor to someone else seeing you move) This is the sum of the sensor lag and the delays imposed by the network; since different network nodes may use different tracking systems, what you really mean here is the network lag. >--lock lag (the time from trying to get a "lock" until you actually have >it... These "locks" are needed to keep more than one person from picking >up an object simultaneously, etc.) Similar comments apply. >The above we consider to be derived parameters. They result from the >number of users, packet size, network configuration, objects in the >world, polygons/sec, cpu speed, network bandwidth, etc, etc, etc. If what you're trying to do is evaluate different network topologies, protocols, and so on then you probably want to eliminate most of the factors you just listed (polys/second, cpu speed, number of objects in the world). One factor that is hard to quantify but which has enormous impact on perceived performance is the "burstiness" or variability of lag that you find on large, heavily-loaded networks like the Internet. -- Bernie Roehl University of Waterloo Dept of Electrical and Computer Engineering Mail: broehl@sunee.UWaterloo.ca Voice: (519) 885-1211 x 2607 [work]