evals for 4/25/2000

Aaras P VASA (avasa@cs.ucsd.edu)
Tue, 25 Apr 2000 09:05:32 -0700 (PDT)

hello sir
these are the evals

aaras

On the duality of Operating system structures

The main objective of the paper is to show that the Message orientated
system and the Procedure-oriented system are duals of each other
Thus the choice about which system to adopt doesn't lie in the
application
which the systems support but on the substrate i.e. machine architecture
and/or programming environment upon which the system is being built

the paper constructs a canonical model of both the systems .The message
orientated system consists of messages,channels ,ports and message
transmission operations while the procedure orientated system
consists of monitors, locks, condition variables and procedure
identifiers.
the two systems are shown to be duals of each other with facililty in
one
system serving the same function as some other in the other system
The dual programs are logically identical to each other and the
performance of a program in also identical to that of its dual system,
Empirical support is given to support the claims.

This paper helps us in eliminating the conflict of choosing a particular
kind of system which is 'supposed' to be better for that application
it also gives us a lot more freedom in the design process.
but it itself is not without controversy. the canonical models do not
map well to the real systems .i think that all real systems need not map
to one of these two systems.
also i think it is not easy to measure the performance of the system
accurately.


The Structuring of System using Upcalls

this paper deals with the idea of replacing asynchronous interprocess
communication between layers with procedure calls .
these procedure calls feature are known as upcalls . The flow of these
calls maps to the natural flow irrespective whether it is up,down or
sideways
The layers themselves are organised as multi-task modules i.e.a
collection of subroutines in different tasks that collaborate
through shared state
shared memory is used with acces mediated by monitor locks.

The swift system is based on these concepts.
it is implemented in a single address space. memory management
and task scheduling for the swift system are designed to meet the
needs of the system

this is a well written paper which clearly shows us both the advantages
and problems with upcalls. upcalls are efficient and simple and help
in eliminating buffering between layers.
but since they do not follow the traditional top down appreoach
they are prone to failures due to some problem in the upper layers
and the problem of indirect recursive call. although techniques to
solve these problems are discussed they are not error free.
a lot of performance in this system depends on the programmer
which might not be such a great thing.
overall with competant programmers upcalls are a definite advantage