[email protected] (Thomas Berg) writes: > Not I. If I compare a typical 3270-interface and a typical > PC/WEB-interface I generally can observe that the response times is > about 50 times better in the 3270-interface. It's also generally less > "cluttered" and easier to handle. > OTOH a typical gui interface can often have more and more > advanced/modern functions etc. And the possibility to display much > more information at one time in a relatively more readable format. > > This is from my experiences from both my employers intranet and from > internet and PC applications in general.
there were lots of studies in the 70s about the benefits of .25sec system response or better (increased human productivity with .25sec or better responses). one of the big issues was while this was possible with 3272/3277 direct channel attached (with some of my carefully crafted mainframe operating systems) ... the newer 3274/3278 direct channel attach hardware latency (best possible case for all 3274s) made it impossible (to achieve .25sec or better). past post with old-time measurements http://www.garlic.com/~lynn/2001m.html#19 when we complained to the 3274 product administrator, the eventualy responses was 3274/3278 wasn't designed for online interactive ... but for "data entry" (i.e. online keypunch). for the last decade or so, i frequently queue up a couple hundred webpages in background browser tabs ... and then only have to deal with purely local latency ... as an aside we had been called in to consult with a small client/server startup that wanted to do payment transactions on their server ... they had also invented this technology they called "SSL" they wanted to use, the result is now frequently called electronic commerce. part of the effort was something called payment gateway ... which handled payment transaction communication between webservers and the internet and payment networks (typically with large mainframes in backend). we gathered elapsed time round trips (from webserver out over the internet to the payment gateway ... through the payment network to the mainframe backends and back). This frequently was around .3secs elapsed round-trip for pure transaction level stuff (at the client browser it could be longer since there is both processing at the webserver as well as another round-trip over the internet). Frequently "internet" slowdowns aren't the actual internet ... but heavy loads on webservers. The big cloud operators have done a lot with significant over provisioning of huge numbers of "on-demand" servers that can be brought on instantaneously to minimize latency due to server load. There is still an issue with larger transmissions around the use of slow-start for contention avoidance. Google is trying to push through standard for much more efficient mechanism ... I have a lengthier discussion in (linkedin) IETF (internet standards) group http://lnkd.in/FCwpMR part of the issue is using rate-based pacing as alternative to slow-start as congestion avoidance ... something we were doing 30yrs ago. -- virtualization experience starting Jan1968, online at home since Mar1970 ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
