Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!elroy.jpl.nasa.gov!ncar!gatech!mcnc!uvaarpa!haven!mimsy!nocusuhs!nmrdc1!minixug!uwalt!waltje
From: waltje@uwalt.nl.mugnet.org (Fred 'The Rebel' van Kempen)
Newsgroups: comp.os.minix
Subject: Re: Amoeba Availability for Minix?
Message-ID: <9103211912@uwalt.nl.mugnet.org>
Date: 21 Mar 91 16:32:13 GMT
References: <1991Mar14.133523.45833@vaxb.acs.unt.edu>
Organization: MicroWalt Corporation, for MINIX Software Development
Lines: 81

kenc@vaxb.acs.unt.edu () wrote:
> What's this rumor that Amoeba is being ported?  I've references to it twice
> today, and it's got me REAL curious.
$ amoeba
amoeba-for-minix: No such rumour or project.

>   I thought that Amoeba was a generic distributed processing-type environment. 
Correct.

> Does this mean that, (assuming we can hook 'em together), several different
> makes of computers...(Mac,Atari,Amiga,Mess-Dos)...could share in computational
> tasks on a large scale?  Is the code transportable?  (Is it available at all?
> grin)
If you have a room full of _large_ machines (4 to 8 Sun's at least), you
can use Amoeba OS.

>   I've ftped the papers Andy, I'll be getting the book on it today.  This
> sounds (so far) like a Good Thing.  I'm wondering...what are the prices of each
> of the nodes on your test setup?  (More specifically, would it be feasible with
> mondo-cheap processors?  Got to start thinking of good projects...directed
> studies are *also* Good Things.)  Heck, for that matter, would cheap processors
You need at least 4 to 8 machines with an iNTEL 386 or Motorola 68030,
one machine with 16M RAM (Bullet File Server), a fast Ethernet, and some
workstations.  Not fun, if you are "just playing around" ...

> be worth putting together, or are higher-power processors required?  (I'm
> thinking 6502/8086/possibly 68000.
Forget it.

>   What kind of resource drain would this put on a system?  Think it'd be
> possible to run this at 115K baud, or does it just _demand_ ethernet?
115K?  You must be joking.  Nope, Ethernet is the least. In Amoeba OS,
_all_ is done using the Ethernet. So....

I think this misunderstanding is caused by the term "amoeba" in context
with MINIX discussions... let me try to clear up this foggy situation :-)

MINIX support several forms of networking:

1. Serial line communications.
   This can be login/getty, data communications using programs like
   term(1), xcomm(1) or kermit(1), possibly with file transfer support
   like Zmodem.

2. Ethernet.
   MINIX has a driver for the Western Digital Etherplus card (WD8003),
   so it is possible to write programs that use it.  To make life sim-
   pler (network programming _can_ be a real pain in the *ss!), Andy
   supplied a simplified version of the RPC protocol that is used in
   the Amoeba distributed OS.  This RPC is easy to use (only 4 user-
   callable routines!), _fast_ (exteremely fast, I might say), and,
   what is also very nice, Amoeba RPC can do "networking" with the
   client and the server on the SAME machine.  This allows for network
   experiments, even when there's no Ethernet around...

So, when MINIX is being discussed, we talk about "Amoeba", but we
actually mean "the Amoeba OS RPC protocol".  This caused the above
confusion :-)

I have done _a_lot_ with Amoeba RPC the past years, and I like it a
lot.  I wrote file servers, name servers (for the MUGNET router),
desktop interfaces, window servers, the lot, all of them used the
Amoeba RPC for some reason....

Since Amoeba is actually very system-independent, I thought it should
be running as a separate server process (AMOEBA), instead of running
in kernel space.  This cleans the kernel from lots of awful hacks,
and it will make the stuff available for Motorola systems.
This Transaction Server will only use Ethernet if it can find an
Ethernet device driver (which, of course, _is_ in kernel space).
The new Ethernet driver is purely message-based, so people can write
their own network protocol servers on top of it.  Alternatively, we
might be able to write a serial link protocol (similar to SLIP) that
can be used to form a slow Amoeba RPC network.....

Oh yeah: all AMOEBA system calls (yeah, all 3 of 'em) now go straight
to AMOEBA; MM is no longer used as a system call interfacer...

Fred.
--
MicroWalt Corporation, for MINIX Development	waltje@uwalt.nl.mugnet.org
Tel (+31) 252 230 205, Hoefbladhof  27, 2215 DV  VOORHOUT, The Netherlands