Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!caip!im4u!ut-sally!pyramid!pesnta!valid!gelfand
From: gelfand@valid.UUCP (Brooks Gelfand)
Newsgroups: net.followup
Subject: Re: FYI: VM systems on the net
Message-ID: <530@valid.UUCP>
Date: Fri, 1-Aug-86 13:21:16 EDT
Article-I.D.: valid.530
Posted: Fri Aug  1 13:21:16 1986
Date-Received: Sat, 2-Aug-86 10:33:18 EDT
References: <455@valid.UUCP> <653@glasgow.glasgow.UUCP> <494@sunybcs.UUCP> <503@valid.UUCP> <281@fai.UUCP>
Organization: Valid Logic, San Jose, CA
Lines: 48

> In article <503@valid.UUCP> gelfand@valid.UUCP (Brooks Gelfand) writes:
> >No operating system, by itself, can imporve hardware reliability -
> >[..] When the
> >hardware crashes you will loose work in progress - the files in core.
> >However, you should not have lost any files on disk unless you were 
> >writing to them at the time of the crash. Or did you perhaps have a
> >disk head crash?
> >
> >Brooks Gelfand
> 
> Not necessarily.  Depends, for instance, on how often the disk is
> updated with directory information.  One of the first things OS
> engineers discover is that keeping directory info in ram decreases
> overhead, speeds things up.  A CPU crash, not a head crash, then
> causes loss of data because the directory (superblock, in Unix terminology)
> has old information.  How well the OS can recover from this is a
> good measure of robustness.  Some do pretty badly.
> 
> Besides, I don't think that was the Col.'s point.  I'm not sure how
> virtual machines are implemented, but it occurs to me that there is
> probably even more possibility of losing files (whole environments?)
> if the hardware (on which several virtual machines exist) gets sick.
> 
> 
> 				Ron
> -- 
> --
> 		Ronald O. Christian (Fujitsu America Inc., San Jose, Calif.)
> 		seismo!amdahl!fai!ronc  -or-   ihnp4!pesnta!fai!ronc
> 
> Oliver's law of assumed responsibility:
> 	"If you are seen fixing it, you will be blamed for breaking it."

Ron, you are correct. VM keeps the disk directory in core when the
disk is accessed. I think I covered that by saying that the file that
is being writen to would be lost. As an example: I am updating File A.
As i write to file A its directory entry (in core is modified). The 
directory entry of all other files remain unchanged. The CPU crashes;
the in core directory is lost. When the CPU comes back up and the 
disk is accessed, the directory in core is the old directory
previous to any modifications to file A. Some blocks in file A
may be modified depending on the program that is modifying the file.
Any blocks added to file A are back on the free list. This is
how CMS would treat the file. Since it is possible to run many different
operating systems under CP (VM), it is impossible to say what happened
without `knowing the exact circumstances.

Brooks Gelfand