Distributed Systems

Delhi Technological University

Introduction
Instructor: Divyashikha Sethia
[email protected]
sites.google.com/site/divyashikhasethia
Divyashikha Sethia (DTU) 1
 Course Objective

Understanding of the issues involved in distributed computer systems and to

investigate the fundamental characteristics of distributed computing systems,
including their models, architectures and designs that exploit rapidly evolving
technology.

Divyashikha Sethia (DTU) 2

 Course Introduction

Prerequisites:

- Operating System
- Computer Networks

Divyashikha Sethia (DTU) 3

 Course Outline
• Unit1:
- Introduction: Introduction to Distributed Systems, Design Goals, Types of
Distributed systems
- Architecture: system architectures and fundamental models, middleware, Threads,
virtualization, client-Server Model, Code migration
• Unit2: Inter-process communication: Communication fundamentals, Remote
Procedure Call, message oriented communication, and stream oriented
communication, multicast communication
• Unit3:

- Time Synchronization: clock synchronization, logical clocks,

- Resource Allocation Synchronization: mutual exclusion algorithms: centralized,
decentralized, distributed and token ring algorithms, election algorithms.

Divyashikha Sethia (DTU) 4

 Course
Unit 4: Replication and Consistency
Outline
need for replication, consistency models: data centric and client
centric consistency models, replica management, consistency protocols:
continuous, primary-based, replicated-write and cache-coherence
protocols.
Unit 5: Fault Tolerance
basic concepts and failure models, process resilience, reliable client-
server and group communication, distributed commit recovery
mechanisms
Unit 6:
-Security in Distributed Systems
- Security in distributed systems, secure channels, authentication, integrity
and confidentiality, access control, security management.
- Naming: Flat naming approaches, structured naming, name space and
resolution, attribute- based naming, directory services, LDAP, decentralized
implementations.
- File Systems
–Distributed file services, example file systems
Divyashikha Sethia (DTU) 5
 Text
Text Books:

Distributed Systems: Principles and Paradigms, 2nd Ed., Andrew S. Taenbaum and
Maarten Van Steen, Prentice Hall, 2007.

References:

• Distributed Operating Systems: Concepts and Design, P.K.Sihna, PHI, 2007.

• Distributed Operating Systems and Algorithms. R. Chow, T. Johnson. Addison-
Wesley Publishing Company, 1997. ISBN 0-201-49838-3.
• Distributed Systems: Concepts and Design, 4th Ed by Coulouris, G, Dollimore, J.,
and Kindberg, T., Addison-Wesley, 2006.

Divyashikha Sethia (DTU) 6

Theory Evaluation

7
Divyashikha Sethia (DTU)
Practical Evaluation

8
Divyashikha Sethia (DTU)
 Definition of a Distributed System (1)

•A distributed system is:

A collection of independent computers that appears to its users as a
single coherent system.
• Utilizes the power of :
- Advancement in computing power of microprocessor
- High speed network communication between computers

9
Divyashikha Sethia (DTU)
 Some uses..

Pool of processors dynamically assigned to users to execute a job in the best

•

manner possible

•Large bank with branch offices over world, each office with:
- Master computer to store local accounts and handle local transactions
- Ability to communicate with other branch offices and central computer

- Transactions can be made without regard to where customer account is

located.

10
Divyashikha Sethia (DTU)
 Definition of a Distributed System (2)

• Middleware to support heterogeneous computers and networks and offering a single

system overview.
• Middleware layer extends over multiple machines.

• Applications distributed over computers.

Divyashikha Sethia (DTU) 11

 Goals

• Resource Availability

• Distribution Transparency

• Openness

• Scalability

Divyashikha Sethia (DTU) 12

 Resource Availability

• Access to Remote resources by all users eg: printers, computers, storage

• Helps collaborate and exchange information

- exchanging files, mail. documents, audio, and video
- electronic commerce

• Threat to Security

Divyashikha Sethia (DTU) 13

Transparency in a Distributed System

Different forms of transparency in a distributed system.

Divyashikha Sethia (DTU) 14
 Openness

Offers services according to standard rules in the form of interfaces

•

described by Interface Definition Languages (IDL)

• Allows an arbitrary process to talk to another process

• Extensible: add new components or replace old ones

Divyashikha Sethia (DTU) 15

 Scalability

• Add more users and resources

Geographical scalability wrt to distribution of users resources across

•

different places

• Administrative scalability

Divyashikha Sethia (DTU) 16

 Scalability Problems when users increase

Concept Example

Centralized services A single server for all users

Centralized data A single on-line telephone book

Centralized algorithms Doing routing based on complete information

Examples of scalability limitations.

Divyashikha Sethia (DTU) 17

 Centralized algorithm

Collect information about load on machines and lines and compute

•

optimal routes for communication.

Collecting and Transporting information to centralized server and then

•

distributing information after analysis overloads network.

• Distributed algorithms characteristics:

- No machine with complete information about system state

- Machine decision based on local information only

- Algorithm not ruined by failure of a single machine

- Global clock

Divyashikha Sethia (DTU) 18

 Geographic scalability

Client server architecture which requires synchronous communication

•

works well on a LAN.

- client, blocks until a reply is sent back

• Communication delays in WAN for response can be longer

•Unreliability of communication over WAN

- local-area networks generally provide highly reliable communication
facilities based on broadcasting

Divyashikha Sethia (DTU) 19

 Administrative scalability

•Scaling across multiple independent administrative domains resolving

conflicting policies for resource usage, management, security

• Security issues across domains

Divyashikha Sethia (DTU) 20

 Scaling Techniques

1. Hiding Communication latencies

1. Distribution

1. Replication

Divyashikha Sethia (DTU) 21

 Scaling Techniques – Hiding
Communication latencies

•Avoid waiting for responses to remote service requests (asynchronous

communication)

• Reply invokes interrupt to complete the previous request

•Alternative to start a new thread of control to perform request – only single

thread blocked for reply while others are in process

• Not suitable for interactive applications requiring quick response time.

eg: accessing databases using forms
• Solution: interactive form filling to be done at client side rather than server

side
Divyashikha Sethia (DTU) 22
 Hiding Communication latencies(2)

1.4

The difference between letting:

a) a server or b) a client check forms as they are being filled

Divyashikha Sethia (DTU) 23

 Scaling Techniques – Hiding
Communication latencies (3)

• Server may check for syntactic errors before accepting an entry

•Better solution: code for filling in form, and possibly checking entries, to the
client, and have the client return a completed form
Eg: Web in the form of Java applets and Javascript

Divyashikha Sethia (DTU) 24

 Scaling Techniques –Distribution (1)

•Splitting components into smaller parts and spreading them

across system

Eg: Domain Name System name space is hierarchically

•

organized as tree of domains, which are divided into

nonoverlapping zones,

Divyashikha Sethia (DTU) 25

Scaling Techniques –Distribution (2)

1.5

An example of dividing the DNS name space into zones and

distributing the naming service
Resolving “nl. vu.cs.flits”: Traverse Z3,Z2,Z1
26
Divyashikha Sethia (DTU)
 Scaling Techniques –Distribution (3)
World Wide Web:

Web is physically distributed across a large number of servers, each handling a

•

number of Web documents.

The name of the server handling a document is encoded into that document's
•

URL.

•It is only because of this distribution of documents that the Web has been
capable of scaling to its current size

27
Divyashikha Sethia (DTU)
 Scaling Techniques - Replication
•Replicate components across distributed system – increasing availability
and balancing load between components for better performance
eg: Geographically close copy can hide communication latency

•Caching is a form of replication – copy of resource in the proximity of the

client accessing the resource
- Difference from replication: on-demand rather than planned

• Leads to consistency problems

- update must be immediately be propagated to all other copies
- Two concurrent updates must be made in the same order.

• Requires global synchronization

Divyashikha Sethia (DTU) 28

 Scalability Conclusion

• Size scalability is the least problematic

• Geographical scalability is a much tougher problem

•Administrative scalability seems to be the most difficult one, because need to

solve nontechnical problems
- peer-to-peer technology demonstrates what can be achieved if end users
simply take over control

Divyashikha Sethia (DTU) 29

 Advantages over Centralized systems

• Resource utilization provides better performance for the price.

• Provide more computing power compared to centralized mainframe

•Computer supported cooperative work involving spatially separated

machines

• Higher reliability

• Incremental growth in computing power

Divyashikha Sethia (DTU) 30

Advantages over PC

Divyashikha Sethia (DTU) 31

Disadvantages of DS

Divyashikha Sethia (DTU) 32

 Classification

• SISD: Single Instruction and Single stream (traditional computer)

SIMD: Single Instruction and Multiple Stream (Array of processors

•

with one instruction to be processed on multiple data handled by

multiple data units in parallel) eg: supercomputers

•MIMD: Multiple instruction and multiple streams – group of

independent computers each with its program counter, program and
data. (Distributed Systems)

Divyashikha Sethia (DTU) 33

 MIMD Classification

Less delay high High delay low

data rate - Parallel data rate -
Distributed

Wires from
Single medium
machine to
connecting all
machine
machines

34
 Hardware Concepts

Different basic organizations and memories in distributed computer systems

Divyashikha Sethia (DTU) 35

 Multiprocessors (1)

A bus-based multiprocessor.

• Bus has address lines, data lines and control lines in parallel.
• Single coherent memory ( memory written is readable to others

coherently after a minute delay)

• Cache memory to increase performance

• Write through cache for uniformity of caches (i.e word written to

cache is written to memory as well so that new CPU accessing it gets

updated value)
• Limited can have at most 64 CPUs
36
Divyashikha Sethia (DTU)
 Multiprocessors (3)

•Write through cache for uniformity of caches (i.e word written to

cache is written to memory as well so that new CPU accessing it gets
updated value)

Snoopy cache – cache constantly monitors the bus for any write
•

occurring to a memory address that is in the cache and updates it.

37
Divyashikha Sethia (DTU)
 Multiprocessors (2)

1.8

a)A crossbar switch

b)An omega switching network

38
Divyashikha Sethia (DTU)
 Multiprocessor
• Crossbar Switch:
- Every intersection between CPU and memory is a physical switch that
can be opened or closed
- Many CPUs can access memory at the same time provided memory

locations are different

- Disadvantage – n CPU, n memory require switches

• Omega network:
- Contains 2x2 switch

- n CPU n memory location require

- Delay due to switching stages

Conclusion: tightly coupled, shared memory multiprocessor is

-

difficult and expensive

39
Divyashikha Sethia (DTU)
 Bus Based Multicomputer

• Each computer has own memory and communicate across bus

• Since traffic is less compared to multiprocessor bus can be a lower speed

LAN

40
Divyashikha Sethia (DTU)
Homogeneous Multicomputer Systems

1-9

a) Grid b) Hypercube

41
Divyashikha Sethia (DTU)
 Types of Distributed Systems

• Distributed Computing System

- used for high-performance computing tasks

• Distributed Information System –

- interoperatabilty & communication of network applications in organization

• Distributed Pervasive Systems

42
Divyashikha Sethia (DTU)
Distributed Computing System(1)

i) Cluster computing
- underlying hardware consists of similar workstations closely
connected by
means of a highspeed LAN
- each node runs same OS

ii) Grid computing

-consists of distributed systems often constructed as federation
of computer systems, where each system may fall under different
Administrative domain, and may have different hardware, software, and
deployed network technology.

43
Divyashikha Sethia (DTU)
 Cluster Computing

•Build supercomputer by simply hooking up collection of simple

computers in high-speed network.

•Cluster computing is used for parallel programming in which single

(compute intensive) program is run in parallel on multiple machines
44
Divyashikha Sethia (DTU)
 Linux Cluster
•Each cluster consists of collection of compute nodes controlled and
accessed by master node that:
- allocates nodes to particular parallel program
- maintains a batch queue of submitted jobs
- provides interface for users of system.

•Master runs middleware for execution of programs and management of

cluster, while compute nodes only need a standard OS.

•Middleware has libraries for executing parallel programs and effectively

provides advanced message-based communication facilities, handling
faulty processes, security, etc.

45
Divyashikha Sethia (DTU)
 Grid Computing
• Computing systems have a high degree of heterogeneity

•Resources from different organizations are brought together to allow the

collaboration

Software provides access to resources from different administrative

•

domains, users and applications that belong to a specific virtual

organization

46
Divyashikha Sethia (DTU)
 Layered architecture for
grid computing

• Fabric layer
- Provides interfaces to local resources at a specific site.
- Allows sharing of resources within a virtual organization

• Connectivity Layer
- Communication protocol for grid transactions for usage of multiple resources
- Security protocols to authenticate users and resources
47
Divyashikha Sethia (DTU)
 Layered architecture for
grid computing

Middleware
Layer

• Resource Layer
- Responsible for managing a single resource
- Functions for obtaining configuration information on a specific resource, perform
specific operations such as creating a process or reading data
• Collective layer
- Access multiple resources
- Services for resource discovery, allocation & scheduling tasks onto multiple
resources, data replication
• Application layer
- Applications that operate within virtual organization &which use grid computing
48
Divyashikha Sethia (DTU)
 Difference between Cluster and
Grid computing
Characteristics of Grid Computing
• Loosely coupled (Decentralization)
Diversity and Dynamism
Distributed Job Management & scheduling

Characteristics of Cluster computing

• Tightly coupled systems
Single system image
Centralized Job management & scheduling system

49
Divyashikha Sethia (DTU)
 Distributed vs Cloud Computing
•Distributed computing/distributed system involve breaking up a
problem which can be solved by a group of computers working at the
same time.

•Cloud computing usually refers to providing a service via the internet.

That service can be pretty much anything, from business software that
is accessed via the web to off-site storage or computing resources.

Divyashikha Sethia (DTU) 50

 Cloud vs Grid Computing
•Grid computing:
-Used in environments where users make few but large allocation
requests. Eg: lab may have 1000 node cluster and users make
allocations for all 1000, or 500, or 200, etc.
- only a few of these allocations can be serviced at a time and others
need to be scheduled for when resources are released
-results in sophisticated batch job scheduling algorithms of parallel
computations.

• Cloud computing:
- lots of small allocation requests. The Amazon EC2 accounts are

limited to 20 servers each by default and lots and lots of users

allocate up to 20 servers out of the pool of many thousands of servers
at Amazon.
- Allocations are real-time and there is no provision for queuing

allocations until someone else releases resourcesSethia (DTU) 51

Divyashikha
 Distributed Information System
Networked applications can be integrated to form enterprise-wide
information system
• Types:

i) Transaction Processing Systems (Database – self study)

- server running application (often including a database) and making it

available to remote programs, called clients.

- clients wrap a number of requests, possibly for different servers,
into a single larger request and have it executed as a distributed
transaction
ii) Enterprise Application Integration
- integration should also take place by letting applications communicate

directly with each other

52
Divyashikha Sethia (DTU)
 Enterprise Application Integration

Integrate applications independent from their databases &communicate

directly
through Middleware in the form of :
i) Remote procedure calls (RPC):
- Application component can effectively send request to another
application component by doing local procedure call, which results in
request sent as message to callee.
- Result sent back and returned to application as result of procedure call
53
Divyashikha Sethia (DTU)
 Enterprise Application Integration

ii) Remote Method Invocation ( RMI)

- allow calls to remote objects instead of application

• Disadvantage: caller and callee need to be up and running at time of

communication

54
Divyashikha Sethia (DTU)
Message Oriented middleware (MOM)
• Disadvantage of RPC & RPI:
- caller and callee need to be up and running at time of communication
- need to know exactly how to refer to each other

• Message Oriented Middleware (MOM)

- applications send messages to logical contact points, described by means

of subject and specifying type of message.

- Middleware takes care those messages are delivered to applications.

- Publish/subscribe systems form an important and expanding

class of distributed systems

55
Divyashikha Sethia (DTU)
 Distributed Pervasive System
⮚Small, battery-powered, mobile, devices with only a wireless connection
hence not very stable
⮚ Configured by owners, but should automatically discover environment

⮚Requirements:

1. Embrace contextual changes.

• Device aware of environmental changes
• Network disconnectivity :automatically connect to another network/
appropriate actions
2. Encourage ad hoc composition.
• Easy to configuration of applications
3. Recognize sharing as the default.
• Easily read, store, manage, and share information
• Efficient discovery of services

56
Divyashikha Sethia (DTU)
 Example of pervasive systems

Home Systems:
⮚

• PCs, gaming devices, smart phones, surveillance cameras, clocks,

kitchen appliances hooked to distributed system
• Self configuring and self managing
• Universal Plug and Play (UPnP) standards by which devices
automatically obtain IP addresses, can discover each other

Electronic Health Care System

⮚

• Devices to monitor well-being of individuals and to automatically

contact physicians when needed and avoid hospitalization
• Various sensors organized in a (preferably wireless) body-area
network (BAN) minimally hindering a person, operating while person
is on the move
57
Divyashikha Sethia (DTU)
 Pervasive Electronic Health Care

Monitoring a person in a pervasive electronic health care system:

(a) Local hub collects data as needed and offloads it to external

storage device time to time.
(b) Continuous wireless connection hooked up to external storage

58
Divyashikha Sethia (DTU)
 Sensor Networks
⮚Used for processing information and form the basis for many medium-
scale distributed systems

⮚Consists of tens to hundreds or thousands of relatively small nodes, each

equipped with a sensing device, battery operated

Distributed Database
⮚

59
Divyashikha Sethia (DTU)
 Sensor Network database

Storing and processing data:

(a) only at the operator's site or (b) only at the sensors. 60
Divyashikha Sethia (DTU)
 Sensor Network database

a) Sensors do not cooperate but simply send their data to centralized

database located at operator's site
• may waste network resources and energy
b) Forward queries to relevant sensors and let each compute an answer,
requiring operator to sensibly aggregate returned answers
• wasteful as it discards aggregation capabilities of sensors which
would allow much less data to be returned to operator

61
Divyashikha Sethia (DTU)
 In-network processing
⮚ Forward query to all sensor nodes along a tree encompassing all nodes
and subsequently aggregate results as they are propagated back to root,
where initiator is located

⮚ Aggregation will take place where two or more branches of the tree
come to together

⮚ TinyDB, implements a declarative(database) interface to wireless

sensor networks
• Use any tree-based routing algorithm.
• Intermediate node collect and aggregate tree results from its children,
along
with its own findings, and send that toward root

62
Divyashikha Sethia (DTU)
 Summary
⮚Distributed systems consist of autonomous computers that work together
to give the appearance of a single coherent system.
⮚Advantage:

• Makes easier to integrate different applications running on different

computers into a single system.
⮚Issues : Resource Availability, Distribution Transparency, Openness,

Scalability
⮚Types of Systems:

• Distributed Computing System, Distributed Information System,

Distributed Pervasive Systems

63
Divyashikha Sethia (DTU)
 Resources

• Distributed Systems - Tanenbaum

• http://www.jatit.org/research/introduction_grid_computing.htm

• http://www.thepicky.com/tech/difference-cloud-computing-vs-grid-

computing/

64
Copy protected with Online-PDF-No-Copy.com Divyashikha Sethia (DTU)