Evaluations for StarOS and Medusa

Yang YU (yyu@cs.ucsd.edu)
Wed, 19 Apr 2000 23:13:08 -0700


General Info about StarOS
StarOS is a object-oriented multiprocessor operation system. It is
designed to make it possible that concurrently
executed processes on diffrent processors coperate with each other.

Emerging of the idea:
Processors are becoming less expensive, thus it is cost-effective to
make cheap processors cooperate with each other.

Hardware that StarOS runs on:
CM*: 50 processors + 3M bytes(1979) main memory
CM* = 5 cluster
cluster = 10 modules+ kmap + map bus
module = LSI-11 processor + memory + Slocal + local bus

StarOs Architecture:
Object- oriented
Object = data | capabilities
Each process has two name spaces : capability name space and the
immediate addressing space
Message based System
Event and mailbox

Software composed of many cooperating, small processes,
StarOS canbe considered as a special Taskforce.
Processes in Taskforce are smaller than those run on uni-processor,
and there are many of them.
Benefits: 1. Cost-performance or absolute performance is better
2. Improve reliability
3. scalability
More communication and synchronization is needed.
Taskforce is an integreted unit.


General Info:
a distributed operating system for CM* mulitmicro processors

Medusa and StarOs are running on the same hardware:
Medusa focus on problems of structure, while StarOS focus on

Goal of Medusa:
How to build distributed operating systems and to exploit the hardware
to produce a system with such attributes:

Hardware influence on medusa
1. CM* is a physically distributed
2. the interprocessor communication facilities

Feature of Medusa:
Control structure of Medusa is distributed
Parallelism is implicit and expected in medusa. All programs are
organized as task forces.

What's in Medusa:
Distributed Structure:
comparison between traditonal approach and distributed
how distribution contributes to modularity and robustness

Taskforce Structure:
Utilities and their implementation(examples of taskforce)
Descriptor Lists

Yang Yu
Department of Computer Science and Engineering
University of California, San Diego