CP SC-662 Distributed Computing Introduction

Introduction to Distributed Systems

• Distributed Systems

• Network OSs vs. distributed OSs

• Research and design issues

• Reading: Coulouris, chapters 1, 2, and 3

Distributed vs. Centralized Systems

• Advantages of Distributed Systems:

– Reliability.
– Sharing of resources.
– Aggregate computing power.
– Openness/Scalability

• Disadvantages of distributed systems:

– Security.
– Physical distribution of resources vs. demand.
– Computing power per node is limited.

1
CP SC-662 Distributed Computing Introduction

Distributed Operating Systems

• Requirements:
– Provide user with convenient virtual computer.
– Hide distribution of resources.
– Mechanisms for protecting resources.
– Secure communication.

• Definition

Distributed OS looks to user like ordinary centralized OS, but runs on

multiple, independent CPUs.
– Use of multiple processors is invisible.
– User views system as virtual uniprocessor.

Distributed vs. Networked OS

• Transparency:
– How aware are users of the fact that multiple computers are
being used?
• Network OS:
– Users are aware where resources are located
– Network OS is built on top of centralized OS.
– Handles interfacing and coordination between local OSs.
• Distributed OS:
– Designed to control and optimize operations and resources in
distributed system.

2
CP SC-662 Distributed Computing Introduction

Network OSs
• Definition:
A network OS is a collection of OSs of computers connected
through a network incorporating modules to provide access
to remote resources.

• Characteristics:
– Each computer has private OS.
– User works on his own machine and remotely logs in to
other computers.
– Users are aware of location of files.
– Limited fault tolerance.

IPC and Transparency

The insider’s view of a centralized OS.
(Roughly patterned after XINU [Comer 1984])

user programs
file system
1 device manager and device drivers
real-time clock manager
2 process coordinator
3 process manager
memory manager
hardware

1 remote files

2 + remote device access

3 + remote process management

3
CP SC-662 Distributed Computing Introduction

A Vanilla Network OS
(Remote Access System [Goscinsky ‘83])

client
(user process)
remote RAS server
requests file
RAS client lib mgmt
rsh lpr ...
local
requests

local OS local OS
remote machine

local machine remote machine

remote resource
Issues:
• Performance! (local and remote)
• Where is the state?
• Serialization of operations.
• Blocking operations

Research and Design Issues

• Communication model
• Paradigms for process interaction
• Transparency
• Heterogeneity
• Autonomy and/or interdependence
• Reliable distributed computing
• Replication

4
CP SC-662 Distributed Computing Introduction

Communication Model
• ISO/OSI Model • An alternative, e.g. Functional, Model
– Physical
– Physical
• same as ISO/OSI
– Datalink
– Datagram
– Network
• connectionless service between
– Transport source and destination process
– Session • location of services
– Presentation – Transport
– Application • reliable transport between client and
server
• “transaction level”
– Binding
• location of resources within the
server
• logical connection between client and
server
– User
• request semantics

Process Interaction: Client/Server

Server: A subsystem that provides a particular type of service to a priori
unknown clients.

• Control functionally distributed among the various servers in the system.

• Control of individual resources is centralized in a server. (localized?)
• Problems:
– Reliability/Availability
– Scalability
– Replication?

5
CP SC-662 Distributed Computing Introduction

Process Interaction: Pipe Model

Pipe: Communication facility to transfer data between processes in
FIFO order. Can be used for synchronization purposes.

• Named/unnamed pipes

• Pipes for secure IPC

• Pipes across network?

• Multicast pipes?

Process Interaction: RPC Model

• Allows a process to call a procedure on a remote computer.
• Looks like a local procedure.
• Blocking characteristics?

• More about this later.

6
CP SC-662 Distributed Computing Introduction

Transparency

Transparency: Make the network invisible to user/applications.

Various degrees of transparency:

• Access Transparency
• Location Transparency
• Name Transparency
• Data Transparency
• Execution Transparency
• Performance Transparency

Autonomy and Interdependence

• Disadvantage generated by interdependence:
– cannot work stand-alone
– globally controlled
– difficult to identify source of authority and responsability
– what about mutual suspicion?
• Reasons for autonomy:
– policy freedom
– robustness
– cooperation between mutually suspicious users