UNIT-I GCC NOTES

UNIT-I
Introduction
Evolutionary Trend in Distributed Computing

The chapter mainly asses the evolutionary change in machine architecture, operating system
platform, network connectivity and application workload.

 To solve the computational problem is centralized computer, parallel and distribution

computing system is used.
 It uses multiple computer to solve large scale problem over internet. The space
computer to solve large scale problem over internet. The space computer sites and
large datacenter must provide high performance computing services to huge no. of
internet uses carefully.
 Based on high demand, high performance computing (HPC) application is no longer
optimal for necessary system performance.
 High Throughput Computing (HTC) system built with parallel and distributed
computing technologies.
 The purpose is to advance network based computing and web services with emerging
new technologies.

Massive Parallel Processors (MPP)

 The HPC & HTC system are both adapted by consumer and high end web scale
computing and information services.
 In HTC system, Peer to Peer network are formed for distributed file sharing and
content delivery application.
 A P2P systems built on many client system cloud computing and web services
platforms are focused on HTC application.

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 1

 UNIT-I GCC NOTES

High Performance Computing (HPC)

HPC systems emphasize the row speed performance. The speeds of HPC system are
improved by the demand from scientific, engineering and manufacturing communities. The
majority of computer users are using desktop computer on large severs when they conduct
internet searches and market drivers computing tasks.

High Throughput Computing (HTC)

HTC systems pays more attention to high flux computing .The main application of high flux
computing is in internet searches and web services. The throughput is defined as number of
tasks completed per unit of time .HTC not only improves bath processing speed ,but also
addresses problems of cost, energy, savings, security and reliability.

Computing paradigms

The Three main computing paradigms

1) Web 2.0 services
2) Internet clouds
3) Internet of Things (IOT)

 With an introduction of SOA, web2.0 service is available.

 Advance in virtualization webs the growth of internet cloud.
 The growth of radio frequency Identification & sensor GPS has triggered the
development of IOT.

A) Centralized computing

This is a computing Paradigm by which all computer resources such as processors,

memory and storage are fully stored and tightly coupled with in one integrated OS.
Eg:-Datacenters and super computer are centralized systems, but they are used in parallel
distributed and cloud computing applications.

B) Parallel computing

All processer are either tightly coupled with centralized shared memory (or) loosely
coupled with distributed memory. It is also accomplished through processing inter processor
communication is accomplished through shared memory via manger passing. A system
capable of parallel computing is known as parallel computer programs shows in parallel
computer are called parallel programs. The process of writing parallel programs is referred as
parallel programming.

C) Distributed computing
A Distributed system consists of multiple autonomous computer .each having its own
private memory communication through a computer network .Information exchange is
accomplished by message passing. A computer program that runs in distributed system is
known as distributed programs .The process of writing distributed programs is known a
distributed programming.

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 2

 UNIT-I GCC NOTES

D) Cloud computing
An internet cloud of resources can either a centralized or distributed computing
systems .The cloud can be parallel or distributed computing .clouds can be built with physical
or virtualized resources ever large data centers.

E) Ubiquitous computing
It refers to computing with pervasive device at any place and time using wired (or)
wireless communication.

F) Internet of Things (IOT)

It is a network connection of everyday objects it is supported by internet clouds to
achieve ubiquitous computing.

Design objectives of HTC & HPC systems

A) Efficiency

 It measures the utilization rate of resources in an execution model by exploiting

massive parallelism in HPC.
 For HTC, efficiency is related to job throughput data access storage and power
efficiency.

B) Dependability

 It measures the reliability and self manager from clip to the system and application
levels.
 The purpose is to provide high throughput services with quality of service (QOS)
even under failure state.
C) Adaption in Processing models

It measures the ability to support large jobs requests ever massive data sets and virtually
cloud resources under various workload and service models.

D) Flexibility in application Deployment

It measures the ability of distributed systems to run both in HPC & HTC application.

Scalable computing trends and Parallelism

Design of parallelism (BLP)

a) Bit level Parallelism (BLP): converts bit-serial processing to world level processing.
b) Instruction level parallelism (ILP): processor executes instructions simultaneously.
c) Data level parallelism (DLP): Through SIMD (simple instruction, multiple data).It
requires more hardware support and compiler assistance.
d)Task level parallelism (TLP): Due to an introduction of multi core processor and clip
multiprocessor (CMP).It is fair due to difficulty in programming and complication of code for
efficient execution on multi core CMPS.
e) Job level Parallelism (JLP): Due to move from parallel processing to distributed
processing. The cases grain parallelism is built on top of fine grain parallelism.

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 3

 UNIT-I GCC NOTES

Application HPC & HTC systems

A) Science and Engineering: Earthquake prediction .Global warming.

B) Business, Education, services: Telecommunication, content delivery, e-communication.
C) Industry and Health care: Banking, stock exchange, hospital automation.
D) Internet & web services Government Application: Data centers, cyber security, online-tax
return processing, social networking.
E) Mission-critical applications: Military command and control, intelligent systems.

Trend towards utility computing

Utility computing

It focuses on a business model in which customer service computing resources from a

paid service provides all grid and cloud platforms are regarded as utility service provides.

HTC in business and HPC in scientific application

Internet of Things

The IOT refer to the networked interconnection of everyday object tools devices .These
things can be large or small and they vary respect to time and place The idea is to every
object using REID(or)sensor(or)electronic technology. The IOT needs to be designed to track
many static (or) morning objects simultaneously .It demands universal addressability of all
the objects. It reduces the complexity of identification search and storage and set the
threshold to filter out fine grain objects.
Three communication patterns exists
a) Human to Human (H2H)
b) Human to Thing (H2T)
c) Thing to Thing (T2T)

The concept behind their communication pattern is to connect things at any time and any
place intelligently .But still it is infancy stage of development.

Cyber Physical System (CPS)

It result of interaction between computational process and the physical world .A CPS
integrates cyber (heterogeneous, asynchronous) with physical (concurrent, information dense)
objects.
It merges 3c technologies.
a) Computation
b) Communication

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 4

 UNIT-I GCC NOTES

c) Control -into an intelligent closed feedback system between the physical world and
information world.
It emphasizes the exploration of virtual Reality application in physical world.

Technologies for Network Based System

a) Multi core CPUs and Multithreading
b) GPU computing
c) Memory storage and wide Area N/w
d) Virtual machine and virtualization Middleware
e) Data center virtualization for cloud computing

a) Multi core CPUs and Multi threading

 Due to growth of component and network technologies the development of HTC and
HPC system are crucial.
 The processor speed is measured in millions of instruction per second (MIPS). The
network bandwidth in measured in MBPS (or) GBPS.
 Each core is essentially a processor with its own private cache (l1 cache) multiple
cases are houses in same clip with L2 cache as it is shared by all cores.
 Multi core and multithreaded CPUs are equipped with high and processors .Each core
can also be multi threaded.

i) Multi core CPUs & many -core GPU Architecture

Multi core CPUs may increase in future but CPU has reached its limit in extending
and exploiting massive DLP (Data Level Parallelism) due to a fore mentioned memory wall
problem
so many-core GPU introduced Both IA-32 & IA-64 instruction set architecture built into
commercial CPUs The GPU has been applied in large clusters to built super computer in
MPPs.

ii) Multi-Threading
Five independent Threads of instruction to four pipelined data paths in five different
processors
a) Four- issue superscalar processor
b) Fine grain multi threaded processor
c) Coarse grain multi threaded processor
d) Dual are CMP

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 5

 UNIT-I GCC NOTES

e) Simultaneous Multi Threaded Processor

The super scalar processer is single threaded with four Functional units. Each of three
multiple threaded processer in four way multi threaded over four functional data paths .The
dual core processer assume two processing cores, each a single threaded two way superscalar
processors.

In superscalar processors same thread is executed .In fine grained switches the execution of
instruction from different thread per cycle. In coarse grained, executes instruction from
different threads simultaneously SMT allows simultaneously scheduling of instruction from
different threads in same cycle. The blank field indicates no instructions.

b) GPU Computing

GPU is a graphics coprocessor (or) acceleration mounted on computer graphics card

(or) video card. A GPU offloads CPU from tedious graphics tasks in video edition
application. These clips can process a minimum of load million polygons per second.

Million polygon per Second

i) Working of GPU

NVIDIA GPU has been upgraded to 128 cores on a single clip. Each core on a GPU
can handle eight threads of instruction. This translates to having up to 1024 threads executed
concurrently on a single GPU.
CPU is optimized for latency cache, while GPU is optimized to delivers high
throughput with explicit management of on-chip memory.
GPUs are designed to handle large no of floating point operations in parallel GPU
offloads the CPU from all data interactive calculation.

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 6

 UNIT-I GCC NOTES

ii) GPU programming Model

CPU is the conventional multi core processor with limited parallelism to exploit. GPU has
many cores architecture that has hundreds of simple processing cores organized as multi
processes.
Each core can have one or more threads. The CPU instructs the GPU to perform massive data
processing. CPU's floating point kernel computation role is largely offloaded to many cores
GPU

c) Memory, Storage and Which Area Networking

1. Memory Technology

The capacity increase of disk array will be greater. Faster processor speed and largely
memory capacity result in a wider gap between processors and memory. The memory will
may become even worse a problem limiting CPU performance.

2. Disks and Storage Technology

The rapid growth of flash memory and Solid State Driver (SSD) also impacts the future of
HPC and HTC systems. Power increases linearly with respect to clock frequency. The clock
rate cannot be increased indefinitely. SSD's are expensive to replace stable disk arrays.

3. System Area Interconnect

The nodes in small clusters are mostly interconnected by an Ethernet switch (or) LAN.LAN
is used to connect client hosts to big servers. SAN connects servers to network storage such
as disk array.NAS connects client hosts directly to disk arrays.

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 7

 UNIT-I GCC NOTES

d) Virtual Machines and Virtualization Middleware

Virtual machines offer novel solutions to underutilized resources, application inflexibility,

software manageability, security in existing physical machines.
The VM can be provisioned for any hardware system. The VM is built with virtual resources
managed by guest OS to run a specific application between VMs and host platform one needs
to deploy a middleware called Virtual Machine Monitor (VMM).
VM1 VM2
----------------------------- Middleware (VMM)
Host platform

a) Bare-metal VM hypervisor, because the hypervisor handles the hardware directly.

b) Host VM Hypervisor, the last OS need not to be modified.

Hypervisor

Bare-metal Hosted

The user application running on its dedicated OS could be bundled together as a Virtual
appliance that can be ported to any hardware platform.

Low Level Virtual Machine Operations:-

a) The VMs can be multiplexed between hardware machines
b) The VMs can be suspended and stored in stable storage
c) A suspended VM can be resumed (or) provisioned to a new hardware platform
d) VM can be migrated from one hardware platform to another

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 8

 UNIT-I GCC NOTES

e) Data Center Virtualization for Cloud Computing

Data Center design emphasizes the performance/price ratio over

speed performance alone. A large data center is built with thousand of servers.
Cost
Servers + disks -- 30%
Chiller -- 33%
UPS -- 18%
A/C -- 9%
Power Distribution -- 7%
-----------
97%

The cloud computing is enabled in four areas. They are:

a) Hardware Virtualization and multi core chips.

b) Utility and grid computing
c) SOA, Web 2.0
d) Autonomic computing and data center automation

System Models for Distributed and Cloud Computing

(OR)
Massive System Classification

The massive systems are classified into four groups. They are
a) Clusters
b) P2P networks
c) Computing grids
d) Internet Clouds

These four system classes consists of millions of computers are participating nodes. These
machines work collectively, collaboratively at various levels.

a) Clusters

 A computing cluster consists of interconnected stand alone computers which work

cooperatively as a single integrated computing resource.
 A cluster of servers interconnected by high bandwidth SAN (or) LAN with shared I/0
devices and disk arrays. The cluster acts as a single computer attached to the internet
via Virtual Private Network (VPN) gateway. The gateway IP address locates the
cluster.
 The system image of computer is decided by the way the OS manages shared cluster
resources. Most clusters have loosely coupled node computers. All resources of server
node are managed by their own OS.
 An SSI is an illusion created by software (or) hardware that presents a collection of
resources as one integrated, powerful resource. It makes the cluster appear like a
single machine to the user.
 It consists of homogeneous nodes with distributed control. Most clusters run on
LINUX OS. The nodes are interconnected by high bandwidth network special cluster
middleware supports are needed to create SSI (or) High availability.

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 9

 UNIT-I GCC NOTES

Cluster Design Issues

1. Complete resource sharing is not available.

2. Without middleware cluster nodes cannot be work together effectively.

b) Computational Grids

The computational grid offers an infrastructure that couples computers, software/hardware,

special instruments, people and sensor together. The computers used in grid are workstations,
servers, clusters and supercomputers. PDAs, laptops can be used as access devices to grid
system.
The grid is built across various IP broadband networks including LAN and WAN already
used by enterprises over internet. The grid integrates computing, communications, contents
and transactions as rented services.
At server end, grid is a network. At the client end, wired (or) Wireless terminal devices.

Grids

Computational P2P
(OR) Data

c) P2P Network (Peer to Peer)

 P2P architecture offers a distributed model of networked systems. In P2P every node
acts as both client and server, providing part of system resources.
 Peer machines are client connected over internet. All client machines act
autonomously to join or leave the system freely. There is no master slave relationship
and no control database.

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 10

 UNIT-I GCC NOTES

 Only the participating peers from the physical network at anytime. The physical
network is simply an adhoc network formed at various internet domain randomly
using TCP/IP protocols.
 Files are distributed in participating peers. Based on communication peer IDs form an
overlay network at logical level. This overlay is a virtual network formed by mapping
each physical machine with ID.
 When new peer joins the system, peer ID is added as node in overlay network. When
a peer removes the system, Peer ID is removed from overlay network

P2P Overlay N/W

Unstructured Structured

 In unstructured overlay network, it is characterized by random graph. There is no

fixed route to send messages or files among nodes. Flooding is applied to send a
query. It results in heavy network traffic and non deterministic search results.
 In structured overlay network follows certain topology and rules for inserting and
removing nodes. Routing mechanisms are also developed and used.

Distributed file sharing of digital

contents (music, videos, etc.)
Collaboration – Skype, MSN
P2P Applications Distributed P2P

Other P2P - .NET, JXTA

ISSUES

1. Too many hardware models and architecture.

2. Incompatibility exists between software and OS.
3. Different network Connections and Protocols.

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 11

 UNIT-I GCC NOTES

d) Internet clouds

Cloud Computing

A cloud is a pool of virtualized computer resources. A cloud can host a variety of different
workloads, including batch style backend jobs and interactive and user facing applications.
Cloud computing applies a virtual platform with elastic resources on demand by provisions
hardware, software and data sets dynamically .Virtualize resources from data centers to form
an Internet cloud for paid users to run their application.

Cloud Service Offerings

a) Infrastructure as a service (IaaS)

b) Platform as a Service (PaaS)
c) Software as a Service (SaaS)

1. Infrastructure as a Service

Put together infrastructure demanded by users. The users can deploy and run on
multiple VM running guest OS on specific applications. The user does not manage or control
the underlying cloud infrastructure but can specify when to request and release the needed
resources.

2. Platform as a Service

This model enables the user to deploy user built applications onto a virtualized cloud
platform. It includes middleware, databases, development tools and runtime support. This
platform includes both hardware and software integrated with specific programming
interfaces. The user is freed from managing the cloud infrastructure.

3. Software as a Service

It is browser initiated application software over paid cloud customers. It applies to business
process, industry applications; ERP on customer side, there is no upfront investment in
servers. On provider side, costs are low, compared with hosting of user application.

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 12

 UNIT-I GCC NOTES

Deployment modes
a) Private
b) Public
c) Managed &
d) Hybrid cloud

Advantages of cloud over Internet

1. Desired location in areas with protected space and higher energy efficiency.
2. Sharing of load capacity among large pool of users.
3. Separation of infrastructure maintenance duties from domain specific application
development.
4. Cost reduction
5. Cloud computing programming and application development
6. Service and discovery
7. Privacy, security, copyright and reliability
8. Service agreements, business models and pricing.

Service Oriented Architecture (SOA)

In grid and web services, Java and CORBA architecture built on traditional OSI layers that
provide base networking abstraction. The base software environment such as .Net, Apache
Axis for web services, Java Virtual Machine for java.
The entity interfaces corresponds to WSDD. Java method and CORBA interface definition
language (IDF). These interfaces are linked with customized, high level communication
system. The communication system supports features including RPC, fault recovery and
specialized routing. These communications built on message oriented middleware
infrastructure such as web sphere (or) Java Message Service (JMS) which provides rich
functionality and support virtualization.

Security is a critical capability that either user or re implements the concepts of IP sec and
secure sockets. Entity communication is supported by higher level services.

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 13

 UNIT-I GCC NOTES

Linked with
Entity Interface Communication System
Customized

Built on

Message Oriented Middleware

a) Web Services and Others

 The web service specifies all aspects of service and its environment. The specification
is carried out using SOAP communication message. The hosting environment then
becomes a universal distributed operating system with fully distributed capability
carried by SOAP messages.
 The REST Approach adopts universal principle and delegates most of difficult
problems to application software. It has minimal information in header, message body
contains all information. It is appropriate for rapid technology and environments.
 In CORBA and Java, the distributed entities are linked with RPC and the simplest
way to build composite applications, to view the entities as objects. For Java writing
java program with method calls replaced by (RMI) Remote Method Invocation, While
CORBA supports similar model with syntax reflecting c++ style of object interfaces.

Differences between Grid & Cloud

A grid system applies static resources while cloud emphasizes elastic resources. Build a grid
out of multiple clouds. So Grid is better than cloud because it explicitly support negotiated
resource allocation.

Elements of Grid Computing

• Resource sharing

– Computers, data, storage, sensors, networks, …

– Sharing always conditional: issues of trust, policy, negotiation, payment etc.,

• Coordinated problem solving

– Beyond client-server: distributed data analysis, computation, collaboration etc.,

• Dynamic, multi-institutional virtual organizations

– Community overlays on classic org structures

– Large or small, static or dynamic

LAYERED ARCHITECTURE OF GRID

• Grid computing is the collection of computer resources from multiple locations

connected together to reach a common goal.

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 14

 UNIT-I GCC NOTES

• The grid can be thought of as a distributed system with non-interactive workloads that
involve a large number of files (data).

• Grid computing is distinguished from conventional high performance computing

systems such as cluster computing in that grid computers have each node set to
perform a different task/application

• Grid computers also tend to be more heterogeneous and geographically dispersed

(thus not physically coupled) than cluster computers.

• Grid sizes are generally quite large .

• Although a single grid can be dedicated to a particular application, commonly a grid

is used for a variety of purposes.

• Grids are often constructed with general-purpose grid middleware software libraries.

There are 5 layers in Grid Architecture

• Fabric Layer

– All physical hardware resources ( storage, sensors, computing servers,

network etc)

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 15

 UNIT-I GCC NOTES

• Connectivity Layer

– Connects resources request to actual resources (establishes communication) –

user permission, data security etc.

• Resource services

– Forming of various resources ( HW and SW) to run an application under

collective operations

• Collective Layer

– Responsible for Monitoring, diagnostics, replication , authorize and

authenticate etc…

• Application Layer

– All application, Control programs , SW dev kits reside here

Standards for GRID environment

• OGSA

– Open Grid Service Architecture

– defines requirements for framework for grid systems that support integration,
virtualization and management.

– Such a framework requires a core set of interfaces, expected behaviours, resource

models bindings.

• OGSI

• OGSA-DAI

• Grid FTP

• WSRF and etc.

GEETHANJALI INSTITUE OF SCIENCE & TECHNOLOGY 16