farhad@corwin.usc.edu (Farhad Khansefid) (03/02/88)
Hi, I read somewhere that the 286 machine in a multiuser environment, sharply degrades in performance when the number of users exceed 3 in normal applications. My questions is whether anyone has used a 386 engine for more than 8 users running data-entry or other types of applications (not heavy calculations) in a xenix environment? - How does the performance degrade compared to a one or two user system? - Did you have to use Smart serial ports? - How far do you think one can push a 386? - Would 28ms disk access time be sufficient or you need to spend some buck for 18/16.5 ms? Any related experience is appreciated. Please send responses directly to me and I will summarize and post. Thanx. farhad@corwin.usc.edu
fox@alice.marlow.reuters.co.uk (Paul Fox) (03/07/88)
In article <7318@oberon.USC.EDU> farhad@corwin.usc.edu (Farhad Khansefid) writes: >I read somewhere that the 286 machine in a multiuser environment, sharply >degrades in performance when the number of users exceed 3 in normal >applications. > >My questions is whether anyone has used a 386 engine for more than 8 users >running data-entry or other types of applications (not heavy calculations) >in a xenix environment? > My experiences of using Xenix/386 are that it seems to perform very well for a single user. For multiple users, it doesnt do too badly as long as you dont push the machine. The hardware I use is a Compaq/386 running at 16Mhz, with a 70MB drive and an Excelan Ethernet card. I use it mostly as a single user, who logs in about 6 times. One thing to be wary of is increasing the disk buffers. By default the system uses about 20% of free memory for buffers. On a 4MB machine, this is about 500K of disk buffers. For large compilations I tried setting it to 1.5MB with pretty disastrous consequences. The machine runs fine until it decides to flush the disk cache. After a long compilation sequence, where most of the disk cache is marked dirty, the system stops for about 10-20seconds whilst it flushes the disk cache. On a single user system, this is unacceptable; for a multi-user system it would be worse. On a multi-user system where most people are running the same programs (eg a database app), where they are not heavily creating files, then increasing the disk buffers can be beneficial, but if one user decides to rebuild the database, or whatever, the other users all suffer. I dont know if this is of any interest. ================= // o All opinions are my own. (O) ( ) The powers that be ... / \_____( ) o \ | /\____\__/ _/_/ _/_/ UUCP: fox@alice.marlow.reuters.co.uk
davidsen@steinmetz.steinmetz.UUCP (William E. Davidsen Jr) (03/09/88)
In article <296@alice.marlow.reuters.co.uk> fox@alice.UUCP (Paul Fox) writes: | [...] | My experiences of using Xenix/386 are that it seems to perform very | well for a single user. For multiple users, it doesnt do too badly as | long as you dont push the machine. | | The hardware I use is a Compaq/386 running at 16Mhz, with a 70MB drive | and an Excelan Ethernet card. How much memory do you have? It sounds like "damned by faint praise" to me, and I wonder if you would give us the rest of the configuration. I have up to three users on a 16MHz 386, and can rarely tell if there is someone else on the machine. I run 1.5MB for the system + 500k per user. When I had less memory, the system did bog down when the third user came on. I plan to go to about 9MB by the end of the year, assuming that memory prices come down. Good points on the disk buffer size. -- bill davidsen (wedu@ge-crd.arpa) {uunet | philabs | seismo}!steinmetz!crdos1!davidsen "Stupidity, like virtue, is its own reward" -me