lisa@phs.UUCP (Jeff Gillette) (10/24/84)
<> Some friends in the humanities computing project at Duke are developing a foreign language authoring system, and would be grateful for the sage advice of netlanders. They are particularly interested in security of the data base (students' records), a UNIX environment, interactive video, and (if possible) voice projection (to pronounce vocabulary/sentences). A possible funding agency has recommended they consider a number of Apollo workstations. Since neither they (nor I) know anything about Apollo, I am asking the experts (you). What is an Apollo system - what kind of hardware/software does it feature, what awesome tricks can it do, how easy would it be for an experienced C/UNIX programmer (on VAXes) to make it show its stuff? And how does the Apollo compare with other competing systems - in processing power, in graphics ability, in user interface, and in cost? Any comments, observations, or information will be appreciated. Send your thoughts to me, and I will forward them to the Duke project. Thanks Jeff Gillette ...!duke!phs!lisa The Divinity School Duke University Durham, NC 27706
holtz@clan.UUCP (Neal Holtz) (10/28/84)
A little over a year ago, we failry carefully evaluated 4 types of workstations, for use in research in engineering software. They were SUN, Apollo, HP9000 and Orcatech (a local company that has currently fallen on hard times). We chose Apollo, and have now had about 9 months experience. We have 3 different models: 1) DN300 - "low end" 1000x800 monochrome, MC68010 based system. 2) DN320 - DN300 + floating point hardware 3) DN660 - "high end" 1000x1000 colour, bit-slice implementation of MC68010 + integrated floating point. (also on network are 2 file servers, and an FPS-164 Attached Processor the latter a fairly high-speed Fortran Engine) We use the standard Apollo O.S. (Aegis) which has the 3 or 4 good ideas of Unix and is cleaned up considerably. Very Unix-like -- anyone with Unix experience should be quite productive in a few hours. They have a version of Unix (AUX) also available, but I don't know it. Performance: ============ Of course, the following are fairly meaningless, but they do give some indication. The hardware that I'm using for comparison is: 1) SUN w.s. Sorry I don't know which model -- one of the earlier ones, but have received several hardware upgrades. 2) Vax 11/750 running 4.2 BSD. 3) Vax 11/780 with Floating point acclerator, running VMS. Numbers are relative to the others - numbers greater than 1 mean the Apollos are faster. All are with respect to measured CPU time. a) matrix multiply - simplest possible Fortran program to initialize and multiply 2 60x60 matrices. Tests combination of floating point and integer (subscript calculations). SUN ? 750 780 DN300 1.1 0.2 0.08 DN320 2.9 0.5 0.2 DN660 7.0 1.5 0.5 b) compiling and linking 2000 line Fortran program. 750 DN300 3.0 DN320 3.0 DN660 6.0+ ? c) Prolog interpretation - answer queries in CHAT database + natural language front end. The same version of Prolog on each. 750 DN300 1.1 DN320 1.1 DN660 3.0+ d) Vector graphics - connect all pairs of points around the circumference of a circle - done so as to minimize floating point computation. SUN ? DN300 0.33 DN660 3.0 e) Numerical integration (monte carlo technique) - 500 line Fortran program for modelling of quarks - many subscript (integer) calculations, as well as the floating point stuff. 780 DN300 0.06 DN320 0.15 DN660 0.5 Summary: performance is good to excellent - particularly when you consider these are personal machines - 750's and 780's tend to be shared. Raw graphics may be slightly faster in comaprably priced SUN's, but the Apollo seems to allow more freedom of choice, and is well integrated into the windowing system. Performance in developing software is much better than a single user 750. Apollo performance is limited more by the number of workstations sharing the disks. Graphics ability ================ Good to excellent, depending on how much you want to pay. The standard monochrome stations, while perhaps not the fastest are quite adequate, and offer other features not usually found (automatic refresh of graphics windows, ability to pan around a large image) User Interface: =============== Very good on the Apollo. The windowing is done right (each foreground process talks to an independent transcript pad that contains the complete transcript since time of creation - processes are controlled simply by typing in the window associated with it). Symbolic debugger is superb (displays source file, with arrow pointing to statement about to be executed, for example). Cost: ===== Unfortunately, Apollo is a little too successful right now. Prices are probably a little (but not excessively) higher than others. Summary: ======== We are very happy with ours. Hardware reliability has not been bad, but it has not been outstandingly good either. Apollo claims they are a software, rather than a hardware company; their software is generally very good. They still have the only functioning local are network (i.e. - processing, files, etc. are completely and transparently distributed). You could do worse.
lisa@UCB-VAX.ARPA (10/30/84)
From: decvax!mcnc!duke!phs!lisa <@MIT-MC.ARPA:decvax!mcnc!duke!phs!lisa@UCB-VAX.ARPA> Some friends in the humanities computing project at Duke are developing a foreign language authoring system, and would be grateful for the sage advice of netlanders. They are particularly interested in security of the data base (students' records), a UNIX environment, interactive video, and (if possible) voice projection (to pronounce vocabulary/sentences). A possible funding agency has recommended they consider a number of Apollo workstations. Since neither they (nor I) know anything about Apollo, I am asking the experts (you). What is an Apollo system - what kind of hardware/software does it feature, what awesome tricks can it do, how easy would it be for an experienced C/UNIX programmer (on VAXes) to make it show its stuff? And how does the Apollo compare with other competing systems - in processing power, in graphics ability, in user interface, and in cost? Any comments, observations, or information will be appreciated. Send your thoughts to me, and I will forward them to the Duke project. Thanks Jeff Gillette ...!duke!phs!lisa The Divinity School Duke University Durham, NC 27706