douglis@cs.vu.nl (Fred Douglis) (02/08/91)
pcg@cs.aber.ac.uk (Piercarlo Grandi) writes: >zs01> Whole file transfer actually works fairly well for an environment >zs01> of single user student workstations. >This is obviously also why Amoeba uses whole file *memory* caching and >even wholly contiguous file allocation on disk. Exactly -- Amoeba requires the whole file to fit in memory, because (1) it simplifies the implementation, (2) it makes RPCs nice and fast, and (3) "all files are small". The problem is, only 99% of files are small. My prediction is that this model has to be thrown away before Amoeba can succeed in a production environment. When a system is used for research only, it's permissible for the system to say "sorry, I can't handle that file for you because it won't fit in my memory," but if a vendor did that they'd be in trouble. One of the biggest problems with this approach is that a file that can be accessed one moment might not be accessible the next time you try. (Pardon me while I bulletproof my windows -- or better yet, fireproof the door. Things could get hot around here :-) P.S. How did this discussion start up in comp.arch? I'm directing followups to comp.os.misc.... (Or trying to, anyway.) -- ============================================================================= Fred Douglis, Vrije Universiteit, douglis@cs.vu.nl +31 20 548-5777 =============================================================================
chet@odin.INS.CWRU.Edu (Chet Ramey) (02/12/91)
In article <AbfmnQy00asVI210Mg@andrew.cmu.edu> zs01+@andrew.cmu.edu (Zalman Stern) writes: >user interface software was called Virtue (I can't remember what it stood >for despite having worked on that software for three years). The best known According to some ITC reports, it stands for `Virtue Is Reached Through Unix and Emacs'. Chet -- Chet Ramey ``There's just no surf in Network Services Group Cleveland, U.S.A. ...'' Case Western Reserve University chet@ins.CWRU.Edu My opinions are just those, and mine alone.