i.1 distributed systems prof. dr. alexander schill dresden technical university computer networks...

16
I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept. http://www.rn.inf.tu-dresden.de

Upload: lawrence-miller

Post on 18-Dec-2015

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.1

Distributed Systems

Prof. Dr. Alexander SchillDresden Technical University

Computer Networks Dept.

http://www.rn.inf.tu-dresden.de

Page 2: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.2Motivation and development tendencies

Desktop PC:

• multitasking

• networking

• direct manipulation, graphical interface

• high performance (CPU, transfer)

• large primary and secondary storage

Areas of application:

• management / development (CASE – Computer Aided Software Engineering)

• team working (CSCW – Computer Supported Collaborative Work)

• group communication

• process control (CIM – Computer Integrated Manufacturing)

Page 3: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.3Sample topology

• networked workstations, also organizationally integrated• super-proportionally increasing communication performance

DistributedOS

LAN (for instance

high performance network)

LAN

L

A

N

L

A

N

WAN

V.A.N.Value added network

Distributed DB

Page 4: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.4Distributed System

• physical computer nodes (processor + storage)• direct / indirect computer coupling

– local networks (Ethernet (CSMA/CD), Token Ring, Token Bus)

– high-performance networks (Gigabit Ethernet, ATM)

– gateways / bridges

– radio networks (GSM, UMTS)

• transport-oriented comm. protocols (TCP/IP, UDP/IP, IPng)• communicating processes

– complete logical connection

– no complete physical connection (communication via inter-components)

• system oriented resources (file system, threads, system programs)• distributed storage, decentralized, co-operative• distributed applications (area specific) on top of distributed systems

Page 5: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.5Example: distributed application

DistributedOS

LAN for instance

high performance network)

LAN

L

A

N

L

A

N

WAN

V.A.N.Value added network

Distributed DB

Logical communication path

Process

Page 6: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.6

Distribution: Purposes

• data, function and load distribution

• decentralization and co-operation

• locality properties and efficiency

• integration of partial applications

• remote resource access

• fault tolerance: reliability and availability

Page 7: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.7

Client(for instance Point of Sale)

Client(for instance Automated

Teller Machine)

Server(for instance

accountserver)

Server(for instance

database)

Requirements:- decentralized system structure- Internet/Intranet-integration- scalability- security concepts- transaction processing- heterogeneity of systems

Client(for instance

Home Access)

Application example

Page 8: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.8Network infrastructure: examples

required bridging betweenheterogeneous networks, system platforms and applications

Middleware

Client(for instance Point of Sale)

Client(for instance Automated

Teller Machine)

Server(for instance

accountserver)

Server(for instance

database)

Client(for instance

Home Access)

X.25

Modem / ISDN / ADSLInternet-access

Fast Ethernet /Token Ring

ATM

ISDN / X.25Fast Ethernet / Token Ring

Page 9: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.9N-tier-architectures

Client(for instance Point of Sale)

Client(for instance Automated

Teller Machine)

Server(for instance

accountserver)

Server(for instance

database)

3-tier: three-level structure; preferable for complex applications

2-tier: two-level structure (user-interface<->Host); simpler, butless flexible

Client(for instance

Home Access)

User interface,if necessarypre-processing (thin client vs.fat client)

Application logic Data management

Page 10: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.10

Middleware and Client/Server: coordination

Client(for instance Point of Sale)

Middleware(for instance

Java RMI, CORBA, .NET, SOAP)

Transport-oriented layers

(for instance TCP/IP, SNA)

Phys. network(for instance Fast Ethernet,

Token Ring, ATM)

Server(for instance account

server)

Middleware

Transport-orientedlayers

Phys. network

Applicationinteraction

Object interaction

Def. of Middleware: “Infrastructure services for distributed applications for bridging of heterogeneity of different systems and networks”

Page 11: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.11

Middleware: Basic technologies

Java (Sun and others):- programming language, applets- Remote Method Invocation (RMI)- Enterprise JavaBeans (EJB): Components

CORBA (Common Object Request Broker Architecture):- object-oriented, language independent; relatively low-level- standard of Object Management Group (OMG)

.NET / COM+ (Component Object Model):- object-oriented, COM+ relatively proprietary, .NET more open- development of Microsoft

Further approaches:- MOM (Message Oriented Middleware)- SOAP (Simple Object Access Protocol), Web Services- transaction monitors, Application Servers

Page 12: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.12

Middleware: general overview

Integrity

Flexibility

Client/Server, Remote Procedure Call (RPC)

(for instance DCE - Distributed

Computing Environment)

Transactionmonitors

Message Oriented Middleware

CORBA-/ RMI-/.NET/SOAP-

Object-orientedBasic comm.

Object TransactionMonitor

Component-Frameworks

(CORBA, EnterpriseJavaBeans, .NET,

WebServices)

ApplicationServer /

Enterprise Appli-cation Integration

Usabilityby

application developer

Page 13: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.13

System models

Client/Server (Remote Procedure Call):

ClientCall

Result

Server

Offeredprocedures

Control thread and data transfer

Separate address spaces

Object-oriented communication:

01 02 06

07

04

03

05Computer 1

Computer 3

Computer 2

Page 14: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.14

Client A

CopyDocument 1

Client B

Client C

CopyDocument 3

Document 1

Document 2

Document n

DocumentServer

. . .

providedocument(“Document 1”)

generateindex(“Document 1”)

providedocument(“Document 3”)

Page 15: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.15

Client-Object A

Client-Object B

Client-Object C

DocumentServer

. . .

D1 = Server.provideDocument (“Document 1”)D1.move (here)D1.<operation>

D2 = Server ...D2.generate Index ()

D3 = Server.provideDocument (“Document 3”)D3.<operation>

Document 1

Document 2

Document n

Document 1

Document 2

Document 3

Page 16: I.1 Distributed Systems Prof. Dr. Alexander Schill Dresden Technical University Computer Networks Dept

I.16Comparison of system models

Client/Server-Model Distributed Object-Oriented Model

processing model procedural model with restrictions

object communication

data access type indirect data access via RPC-Server

direct access to objects

data transfer value parameter semantics

reference parameter semantics

identity not system-wide unique

system-wide unique

granularity server of coarse granularity

objects of any granularity

placement fixed placement modifiable placement

higher transparency grade and improved influence on distributionwith object-oriented model