chap01 prasad

7/28/2019 Chap01 Prasad

1/34

Distributed System: Definit

A distributed system is a piece of software

sures that:

A collection of independent computers

appears to its users as a single coherent s

tem


2/34

Goals of Distributed System

Connecting resources and users

Distribution transparency

Openness

Scalability


3/34

Backgrou

Developing Collaborative applications ovheterogeneous devices and data stores

Autonomous and mobile data stores

Wireless (or wired) networks of various

Devices of varying capabilities (pagers, c

Limitations of Current Technology

Explicit andtedious(dataandnetwork) p


4/34

System on Mobile D

An Integrated Programming and

UniformUniform WebWeb ServiceServiceview of devview of dev

persistent objectpersistent object--viewviewof mobile da

Rapid development of reliable and

high-level programming and deplo

Leverageoff existingserver applic


5/34

SyD Kernel Architecture

SyDDirectory

SyDListener

SyDEngine

S DLi k

3. Regis ter

GloballySyDAppO

Server

Client UI

SyD Kernel

2. Invoke

2. publish

1. Lookup

2

.Lookup

3.Re

SyD

SyD service

intellig

SyD deviceinvoca


6/34

Distribution Transparency

Transparency Description

Access Hides differences in datarepresentation and invocationmechanisms

Location Hides where an object residesMigration Hides from an object the ability o

system to change that objects lo

Relocation Hides from a client the ability of asystem to change the location of

object to which the client is bound


7/34

Degree of Transparency

Observation: Aiming at full distribution trans

may be too much:

Users may be located in different contine

tribution is apparent and not something yto hide

Completely hiding failures of networks an

is (theoretically and practically) impossib


8/34

Openness of Distributed Sys

Open distributed system: Be able to inter

services from other open systems, irrespectiv

underlying environment:

Systems should conform to well-defined in Systems should support portability of

tions

Systems should easily interoperate


9/34

Policies versus Mechanism

Implementing openness: Requires suppor

ferent policies specified by applications and

What level of consistency do we require fo

cached data? Which operations do we allow download

to perform? Which QoS requirements do we adjust in

of varying bandwidth?


10/34

Scale in Distributed System

Observation: Many developers of modern dis

system easily use the adjective scalable with

ing clear why their system actually scales.

Scalability: At least three components:

Number of users and/or processes

(size scalability)


11/34

Techniques for Scaling

Distribution: Partition data and computation

multiple machines:

Move computations to clients (Java apple Decentralized naming services (DNS) Decentralized information systems (WWW

Replication Make copies of data a ailable


12/34

Scaling The Problem

Observation: Applying scaling techniques

except for one thing:

Having multiple copies (cached or replica

leads toinconsistencies: modifying one

makes that copy different from the rest.

Al k i i i t t d


13/34

Distributed Systems:

Hardware Concepts

Multiprocessors

Multicomputers

Networks of Computers


14/34

Multiprocessors and

Multicomputers

Distinguishing features:

Private versus shared memory Bus versus switched interconnection

Shared memory Private memory

M MM M M M


15/34

Networks of Computers

High degree of node heterogeneity:

High-performance parallel systems (mult

sors as well as multicomputers) High-end PCs and workstations (servers) Simple network computers (offer users o

work access) Mobile computers (palmtops, laptops)


16/34

Distributed Systems:

Software Concepts

Distributed operating system

Network operating system

Middleware


17/34

Distributed Operating Syste

Some characteristics:

OS on each computer knows about the ot

puters OS on different computers generally the Services are generally (transparently) dis

across computers


18/34

Multicomputer Operating Sys

Harder than traditional (multiprocessor) O

cause memory is not shared, emphasis shifts

cessor communication by message passing:

Often no simple global communication:

Only bus-based multicomputers provi

ware broadcasting

Effi i t b d ti i t


19/34

Network Operating System


Each computer has its own operating sys

networking facilities Computers work independently (i.e., they

have different operating systems) Services are tied to individual nodes (ftp

WWW)


20/34

Distributed System (Middlew


OS on each computer need not know ab

other computers OS on different computers need not gene

the same Services are generally (transparently) dis

t


21/34

Need for Middleware

Motivation: Too many networked applicatio

hard or difficult to integrate:

Departments are running different NOSs Integration and interoperability only at leve

itive NOS services Need for federated information systems:

C bi i diff d b b


22/34

Middleware Services (1/2

Communication services: Abandon primitive

based message passing in favor of:

Procedure calls across networks Remote-object method invocation Message-queuing systems Advanced communication streams

E ifi i i


23/34

Middleware Services (2/2

Control services: Services giving applicatio

trol over when, where, and how they access d

Distributed transaction processing Code migration

Security services: Services for secure pro

d i i


24/34

Comparison of DOS, NOS

and Middleware

1: Degree of transparency

2: Same operating system on each node?

3: Number of copies of the operating system

4: Basis for communication

5: How are resources managed?

I h l ?


25/34

ClientServer Model

Basic model

Application layering

ClientServer architectures


26/34

Basic ClientServer Model (

Characteristics:

There are processes offering services (se There are processes that use services (c Clients and servers can be distributed ac

ferent machines Clients follow request/reply model with re

using services


27/34

Basic ClientServer Model (

Servers: Generally provide services related to

resource:

Servers for file systems, databases, impl

tion repositories, etc. Servers for shared, linked documents (W

tus Notes) Servers for shared applications


28/34

Application Layering (1/2

Traditional three-layered view:

User-interface layer contains units for an

tions user interface Processing layer contains the functions o

plication, i.e. without specific data Data layer contains the data that a client

manipulate through the application comp


29/34

Application Layering (2/2

Querygenerator

Rankingcomponent

HTMLgenerator

User interface

Keyword expression

Database queries

Ranked listof page titles

HTML pagecontaining list

Ple

Ule


30/34

Client-Server Architecture

Single-tiered: dumb terminal/mainframe con

Two-tiered: client/single server configuration

Three-tiered: each layer on separate machin

Traditional two-tiered configurations:


31/34

Alternative C/S Architectures

Observation: Multi-tiered architectures seem

stitute buzzwords that fail to capture many

clientserver organizations.

Cooperating servers: Service is physicallyuted across a collection of servers:

Traditional multi-tiered architectures Replicated file systems


32/34

Alternative C/S Architectures

Essence: Make distinction between vertical a

izontal distribution

Replicated Web servers eachcontaining the same Web pages

Front endhandlingincomingrequests

Requestshandled inround-robinf

ashion


33/34

r B H !

! !

&

&


34/34

r B F ! C B !

chap01 prasad

Documents