evaluation. Thank you

Qiao XIN (qxin@cs.ucsd.edu)
Thu, 27 Apr 2000 22:14:29 -0700 (PDT)

QIAO XIN

Evaluation of the Paper: Emerald System

This paper discusses the mobility in the Emerald system which is
object-based distributed system. I think the design of the system is
successful in achieving its goal of mobility, distribution and
performance in the local case.

Differing from other existing systems aiming the same goals, Emerald
system has lots of new ideas. Its objects can be small data-only
objects and mobile distributed objects. I think an important and
prominent issue in its design is the call-by-object-reference
parameter passing method. The mobility of the objects enables to avoid
remote references by moving argument objects to the site of a remote
invocation which can be performed by the programmer using
call-by-move. Processes in Emerald are considered as a stack of
activation records and remote invocation is considered as the
activation record moves to the destination node.

In Emerald, all objects on a node share a single address space and
objects are addressed directly. Invocation are implemented through
procedure call or in-line code. Finding the location of a given object
in the distributed system uses the algorithm based on the concept of
forwarding address. However, one result from the above is that
pointers must be translated when an object is moved. The addressing
and implementation structure in Emerald makes the actual moving of
objects straightforward. Emerald's garbage collection scheme is also
good since it allows the collectors to proceed in parallel with
executing processes in the distributed system.

Despite of the benefits of the system, I think we should also notice
the constraints the system relies on, such as appropriate choice of
languages semantics, tight coupling between the compiler and the
kernel and careful attention to implementation. The implementation of
mobility in Emerald involves trade-offs between the performance of
mobility and that of fundamental mechanism, for example, the use of
direct memory addresses increase the performance of local invocations
but increase the cost of mobility.

Above all, I still think Emerald is a good design and it should be
realized without the assumption of local area network of modest number
of node and the nodes are homogeneous.

Evaluation of the Paper: Experience with Grapevine

Grapevine is a distributed, replicated system that provides message
delivery,naming, authentication, resource location and access control
services in an internet of computers. This paper describes the growth
of the system by addressing achievements and problems in their
experience and some approaches to solve the problems.

I think this paper does not have focus and it is very hard for me to talk
about the paper following an obvious clue. Below I will follow the
paper in this evaluation.

Grapevine is designed to meet the principle of distributed system of
increasing capacity by adding servers of a fixed power rather than by
using more powerful servers. The primary method for preventing scaling
problems is division of the registration database into
registries. New ways to handle general interest distribution lists is
added. However, it still has the scale effect depending on the size
of underlying internet.

Grapevine's configuration decision involves (1)large server is needed
to enhance message sharing.(2)The idea of adding a pair of machines of
two kinds servers at time of heavy load is proposed. (3)add
organizational registries in addition to geographical ones. (4) place
a primary inbox in the server and a secondary inbox nearby. (5)many
replicas of registries are need.

Grapevine has some problems related with distributed systems such as
delay in propagating registration database changes, unremoved deleted
names, failure to eliminate a duplicate message from an inbox.

Since Grapevine design decisions depends on the load. Adjusting to the
actual load is necessary. New mechanism is added to propagate one sort
of database update. Service provided for users with terminal and
permit random access to the messages in the inboxes. To become the
source of authentication and access control information for the
internet, Grapevine now takes the approach of caching the result of
the checking and changes the nested group structure.

As to reliability, the technique of replications among several servers
implies the fact that using redundancy to achieve reliability.