[email protected] (Mike Schwab) writes: > I miss my 3278. we had big battle with kingston (before communication group moved to raleigh) about 3278 (& 3274 controller) for interactive computing. eventually they came back and said that 3278 wasn't designed for interactive computing ... but pure data-entry.
part of the issue (compared to 3277/3272) was that much of the electronics had been moved out of the terminal back to the shared controller ... significantly reducing the terminal (3278) manufacturing cost. however, this significantly drove up the coax "chatter" ... because nearly everything required electronics back in the shared controller. There had been some number of hardware tweaks to the 3277 terminal hardware to improve the interactive computing characteristics ... which also became impossible with the migration of the electronics back into the 3274 controller. the additional processing load on the shared 3274 controller (local chanel attach) also significantly drove up its channel busy overhead (in addition to making the 3278 slower operation and unsuitable for interactive computing). Later in the migration to (coax) terminal emulation, a emulated 3277 would have 3-4 times the upload/download throughput than an emulated 3278 (because of the enormous amount of coax protocol chatter to support the paradigm where the terminal electronics are back in the controller). past post with old hardware response measurements compared channel attach 3272 with channel attach 3274 http://www.garlic.com/~lynn/2001m.html#19 3270 protocol as part of human productive improvement studies associated with quarter-second response (along with typical TSO/CMS system response numbers typical of internal corporate datacenters). for a little other topic drift ... STL lab was bursting at the seams and they were moving 300 people from the IMS group to an off-site building. The group had done some tests with "remote 3270s" ... but found the human factors totally intolerable (expecially since they were used to vm370/cms with channel attach controllers). I did support for channel extender (about five miles) that allowed local 3274 channel-attached controllers at the remote site with 300 3270s ... and things were actually better than when they were in the building (but you also got to realize that these were 3278/3274 combination). The 300 from the IMS group found little difference between in-building and off-site (with the channel extender) ... but what happened was the overall throughput of the systems improved by 10-15%. The issue was that it was common to mix 3274 and disk controllers on the same channels. Putting in the channel extenders got all of the 3274 controllers off the same real channels with disk controllers (removing the enormous 3274 channel busy overhead) ... improving disk throughput. -- 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
