Distributed Operating

System
What is a Distributed Operating System ?
 Distributed Operating System

A Distributed Operating System is the one that runs

on multiple, autonomous CPUs which provides its
users an illusion of an ordinary Centralized
Operating System that runs on a Virtual
Uniprocessor.


“If you can tell which computer you are using, you are
not using a distributed operating system.” -
Tanenbaum
 Distributed Operating System

§ Runs on a cluster of machines

§ Users get the feel of a single processor - virtual
uniprocessor
§ Transparency is the driving force.
§ Requires:
 A single global IPC mechanism
A global protection mechanism
Identical process management and system calls at all nodes
Common file system at all nodes
 How is it different ?

The Distributed Operating System is unique and

resides on different CPUs.
User processes can run on any of the CPUs as
allocated by the Distributed Operating System.
Data can be resident on any machine that is the part
of the Distributed System.
All multi-machine systems are not Distributed
Systems.
 Distributed OS vs. Network OS.

ü User is not aware of the üUser is aware of the

multiple CPUs. existence of multiple CPUs.
ü Each machine runs a part üEach machine has its own
of the Distributed
Operating System. private Operating System.
ü The system is fault- üThe system is not fault-
tolerant. tolerant.
 Why Distributed Operating Systems ?

Provides Transparency.

Reliability and Availability.

Price/Performance.

Resource Sharing


Simplicity of Software.

Scalability
 Transparency

Transparency Description

Access Hide differences in data representation and how a resource is accessed

Location Users are not aware of the positioning of the resources in the system

Migration Resources can move without changing names

Relocation Hide that a resource may be moved to another location while in use

Replication Users should not be aware of multiple copies of resource

Concurrency Hide that a resource may be shared by several competitive users

Failure Hide the failure and recovery of a resource

Persistence Hide whether a (software) resource is in memory or on disk

 Reliability

§ Fault Avoidance


§ Fault Tolerance


§ Fault Detection & Recovery

 Scalability

Avoid centralized entities


Avoid centralized algorithms

Perform most operations on client workstations

 Distributed Operating System Models

 Minicomputer Model
ØIt consists of a few minicomputers each with
multiple users.
ØSimple outgrowth of the Central Time-Sharing
Systems.
ØEach user is locally logged-on to one machine and
remotely logged-on to other machines.
Ø(Logged-in Users / Available CPUs) < 1
 Distributed Operating System Models

 Workstation Model
ØEach user has his personal workstation and nearly all
work is done on the workstation.
ØEach user is locally logged-on to one machine and
remotely logged-on to other machines.
ØIt supports single, global file-system that provides
location-independent data access.
Ø(Logged-in Users / Available CPUs) ~ 1
 Distributed Operating System Models

 Processor Pool Model



ØWhen an user needs to perform computation, a

processor is allocated from the processor pool to the
user task.
ØTerminals are diskless terminals and Pool of backend
processors handle processing
Ø(Logged-in Users / Available CPUs) > 1
Distributed Operating System Models

 Client-server Model


ØUser has local workstation

Ø
ØPowerful workstations serve as servers
(file, print, DB servers)

 Problems with Distributed Operating System

Communication Protocol Overhead.

Lack of Simplicity.
High requirement of the degree of fault tolerance.
Lack of global state information (e.g. No global
Process Tables).
Atomic Transactions.
Process and Data Migration (e.g. During Load
Balancing and Paging respectively).
 Design Issues

Communication Primitives
Naming and Protection
Resource Management
Fault-Tolerance
Services
 Communication Primitives

Remote Procedure Call



Parameter Passing: Passing reference parameters over

the network is not easy. A unique system wide pointer for
each object is needed to access it remotely.

Parameter Representation: Incompatible representation
of data across network.

Client-Server Binding: Sometimes it is important to
know the details of the servers while handling RPC calls.
 Naming and Protection

Naming is required for Object Recognition.

 Naming as Mapping
 Problem of mapping between two domains.
Name Servers

Maintain a table or
database of the name-to
object mapping.
Services, processes, etc
need to register with the
underlying naming
system.
 Resource Management

 Managing resources without having accurate global

state information is difficult.
Distributed OS do not have tables that provide up-to-
date status information of all the resources being
managed.
Considerations:

§ Processor Allocation
§ Scheduling
§ Load balancing
§ Distributed Deadlock Detection
 Fault Tolerance

A fault tolerant system is the one that can continue

functioning, perhaps in a degraded form, even if
something goes wrong.

Two radically different approaches:
§ Redundancy Techniques- create multiple copies of a process
and messages sent to a process
§ Atomic Transactions- property to run-to completion or do
nothing
 Services

In a Distributed Operating system, it is useful to have user



level server processes to provide functions that have been

traditionally provided by the operating system leading to
the microkernel approach of the operating system design.


Some of them are:



§ File Service.
§ Print Service.
§ Process Service.
§ Terminal Service.
§ Mail service.
Comparison of some Distributed Operating
Systems
 Thank You