evals for 4/27/2000

Aaras P VASA (avasa@cs.ucsd.edu)
Thu, 27 Apr 2000 10:54:36 -0700 (PDT)

hello sir
sorry i am so late this time. it wont happen again


Experience with Grapevine: the grwth of a distributed system

The objective of this paper is to discuss the performance of
the grapevine system based on their experience i.e. how well the
design has performed in actual operation

grapevine is a distributed system. The grapevine servers
are mainly divided into the message servers and the registration
servers providing the message and registration services.
The registration service is based on a registration database which is
distributed and replicated. Reliability is achieve by replication.

this paper shows well how the grapevine has scaled
some changes have had to be made for this.decreasing the size of update
, allowing login thru graevine servers are some of them
this paper doesn't discuss well how the configuration decisions take place
which could be a problem.
also it doesn't scale for users accesing the message system via
terminal service.

Fine-Grained Mobility in the Emerald System

The main point discussed in this paper is to discuss and implement
fine-grained mobility i.e. object mobility in the Emerald System
such that the peformance is not affected

Emerald is object-based designed for the distributed environment
It uses a single object model with support fro abstract types.
it haslanguage support for object mobility including primitives and
call-by-move and call-by value parameter-passing modes.
It uses shared memory. the processes in emaerald are viewed as
activation records.
direct memory addresses are used, the modification of which increases
the cost of mobility.

i think this is well written paper discussing the problems and their
solutions for implementing object mobility.i think the call-by-move
and call-by-value are alternatives to the called-by-object-
refernce semantic.but they will not be nearly as effective with larger
objects.i think that semantics are definitely useful but their
usefulness is slightly exaggerated in the paper
trade-offs have been made to improve local references. complexity of
the mobility results from this.
also this syytem is not meant to be scalable.
it also expects all nodes to be homogenous something not always