Definition Definition of a Distributed System: A collection of - - PDF document

Lecturer: Hadi Salimi Lecturer: Hadi Salimi Distributed Systems Lab, School of Computer Engineering I U i i f S i d T h l Iran University of Science and Technology, hsalimi@iust.ac.ir

Definition

Definition of a Distributed System:

A collection of independent computers that appears to its

i l h t t users as a single coherent system. Pay attention to the:

Pay attention to the:

Independent computers A set of users Single coherent system

No assumption about: No assumption about:

Computer hardware, operating system, software, etc. Network interconnects

Components

Components of a distributed system

Distributed Systems Goals

Four different goals that should be met to make building

the distributed system worth of effort:

Making resources available to users Hide the fact that resources are distributed (Transparency) It should be open It should be scalable It should be scalable

Transparency

Different types of transparency in a distributed system

Transparency vs. Performance

There are cases in which hiding all aspects of a distributed

system is not a good idea:

Requesting to have your daily newspaper in your mailbox Multiple tries of a client to re‐connect a server in transient

failure cases

Making several replicas consistent Making several replicas consistent Hiding the distribution of a system in pervasive

environment may not be a good idea always

Openness

An open distributed system is a system that offers services

according to standard rules that describe the syntax and semantics of those services.

S i di ib d i f

Some services at a distributed system expose interfaces to

the world.

Interfaces could be defined using Interface Definition

Interfaces could be defined using Interface Definition Language (IDL)

What does an IDL describe? Syntax or semantics?

Policy and Mechanism

There should be a clear border between the definition of a

service and its implementation.

These two parts are called the policy and mechanism As an example, consider the case of web caching policy

d diff f i i l i and different parameters for its implementation.

Scalability

Scalability of a system can be measured at least three

different dimensions:

Size scalability Geographical scalability Administrative scalability

Scalability Problems

Scalability Problems

Scaling Techniques

There are three main techniques for making the systems

scalable:

Hiding communication latencies Distribution Replication

Scaling Techniques (Cont.)

(a) Checking the form in the server side (b) checking the form in ( ) g ( ) g the client side

Scaling Techniques (Cont.)

Distribution technique for making the systems scalable Distribution technique for making the systems scalable

Pitfalls

False assumptions made by first time developer:

• The network is reliable.
• The network is secure.
• The network is homogeneous.
• The topology does not change.
• Latency is zero.

B d idth i i fi it

• Bandwidth is infinite.
• Transport cost is zero.
• There is one administrator

There is one administrator.

Distributed System Types

Types of Distributed System:

Distributed Computing Systems Distributed Information Systems

Enterprise Application Integration (EAI)

Di ib d P i S

Distributed Pervasive Systems

Cluster Computer Systems

Cluster computer systems (Linux‐based Beowulf)

Grid Computing

Grid are different from clusters in:

Gris are highly heterogeneous Grid are distributed geographically

Grids are based on the concept of Virtual Organizations

and sharing them among different users.

Pervasive Systems

Pervasive systems are characterized by:

The environment is not stable Devices are small and power matters Lack of human administration Networks are performed add‐hoc

E l i l d

Examples include:

Sensor networks Electronic Health‐case systems Electronic Health‐case systems Home systems

Spectrum of HW

MPP (Massively Parallel Processors) ‐ Multiprocessor

Expensive, need HP networking, simultaneous access to

shared memory, single address space, hardwired, tightly couples COW (Cluster of Workstations) – Multi‐computer

COW (Cluster of Workstations) Multi computer

Cheaper, easy to make, message passing, not hardwired,

flexible, loosely coupled

Distributed Shared Memory (DSM)

a)

Pages of address space distributed among four machines machines

b)

Situation after CPU 1 references page 10

)

Si i if

c)

Situation if page 10 is read only and replication is used p

Network and Distributed OSs

Distributed Operating System

DOS

Network Operating System

HW HW

Network Operating System

NOS NOS HW HW