sysrick@cfht.hawaii.edu (Rick McGonegal) (04/24/91)
We are taking a lot of flak from our user base lately due to the poor response of our GUI. After looking into it we have found, for our most complicated form, the following statistics: System A: System B: code executes on: HP9000/835 32 Mb RAM same display server HP9000/825 16 Mb RAM, SRX HP9000/340 8 Mb RAM form "popup" time 15 seconds 5 seconds This would seem to indicate a factor of three performance penalty in using an 800 with SRX over a vanilla 340 ?? [the code reads a lot of files each time which explains the 4-5 seconds] Is there any one else out on the net with experience in this area - i.e. running X11 on an SRX display? Is there, perhaps, an X11 display server optimized for the SRX? any help, comments would be appreciated. thanks in advance rick mcgonegal CFHT Corp.
ian@rathe.cs.umn.edu (Ian Hogg) (04/25/91)
In article <sysrick.672447410@cfht.hawaii.edu> sysrick@cfht.hawaii.edu (Rick McGonegal) writes: >We are taking a lot of flak from our user base lately due to the poor response >of our GUI. After looking into it we have found, for our most complicated form, >the following statistics: > > System A: System B: >code executes on: HP9000/835 32 Mb RAM same >display server HP9000/825 16 Mb RAM, SRX HP9000/340 8 Mb RAM >form "popup" time 15 seconds 5 seconds > >This would seem to indicate a factor of three performance penalty in using an >800 with SRX over a vanilla 340 ?? [the code reads a lot of files each time >which explains the 4-5 seconds] > Not necessarily. Since you are using remote displays you have to account for possible delays in the networking. Also, how many users/processes are running on the 825? It is a multi-user or some kind of server machine? >Is there any one else out on the net with experience in this area - i.e. running >X11 on an SRX display? Is there, perhaps, an X11 display server optimized for >the SRX? > >any help, comments would be appreciated. > >thanks in advance >rick mcgonegal >CFHT Corp. -- Ian Hogg email: rathe!ian@cs.umn.edu ...!umn-cs!rathe!ian Rathe, Inc ianhogg@cs.umn.edu 366 Jackson Street phone: (612) 225-1401
jwright@cfht.hawaii.edu (Jim Wright) (04/26/91)
ian@rathe.cs.umn.edu (Ian Hogg) writes: >sysrick@cfht.hawaii.edu (Rick McGonegal) writes: > > System A: System B: > > code executes on: HP9000/835 32 Mb RAM same > > display server HP9000/825 16 Mb RAM, SRX HP9000/340 8 Mb RAM > > form "popup" time 15 seconds 5 seconds > > > > This would seem to indicate a factor of three performance penalty in using > > an 800 with SRX over a vanilla 340 ?? [the code reads a lot of files each > > time which explains the 4-5 seconds] > Not necessarily. Since you are using remote displays you have to account for > possible delays in the networking. Also, how many users/processes are > running on the 825? It is a multi-user or some kind of server machine? All machines are within a single, dedicated subnet. For all machines, the only user logged in is the one doing the benchmarks. Both the 825 and 340 should have roughly the same number and kind of processes executing. The 835 should be in a virtually identical state regardless of where the X display is. -- Jim Wright jwright@cfht.hawaii.edu Canada-France-Hawaii Telescope Corp.
harry@hpcvlx.cv.hp.com (Harry Phinney) (04/29/91)
Rick McGonegal writes: >Is there any one else out on the net with experience in this area - i.e. running >X11 on an SRX display? Is there, perhaps, an X11 display server optimized for >the SRX? Well, yes, we have a fair amount of experience running X on an SRX :-) From the performance comparison you posted I would guess that the 340 has a "CH" (i.e. 98550A) display card. The "CH" display system has several hardware features which are quite useful for the X server, and that the SRX lacks. Given equal levels of software tuning, an X server for a CH will be quite a bit faster than one for an SRX. I am, however, surprised at the ~3x difference you are seeing. It may be that you are hitting an unusually bad spot in the SRX server. Is there any chance you could make available a sample program which demonstrates this level of difference? An executable would be fine. I can't guarantee that we can offer an improvement, but I'm certainly willing to take a further look at the problem. Harry Phinney harry@hp-pcd.cv.hp.com