Module 2.pdf

Module 2
Parallel and Distributed Programming Paradigm

2
Parallel and Distributed Programming Paradigms Introduction, Parallel
and distributed system architectures, Strategies for Developing Parallel
and Distributed Applications, Methodical Design of Parallel and
DistributedAlgorithms
Cloud Software Environments - Google App Engine, Amazon AWS,
Azure Open-Source tool

3
 Computing is the process to complete a given goal-oriented
task by using a computer technology.
 Computing may include the design and development of
software and hardware systems for a broad range of
purposes, often consists of structuring, processing and
managing any kind of information.

Parallel computing systems are the simultaneous execution of the
single task (split up and adapted) on multiple processes in order to
obtain results faster.
The idea is based on the fact that the process of solving a problem usually
can be divided into smaller tasks (divide and conquer), which may be
carried out simultaneously with some coordination .
The term parallel computing architecture sometimes used for a computer
with more than one processor (few to thousand), available for processing.
5

The recent multi-core processors (chips with more than one processor
core ) are some commercial examples which bring parallel computing to
the desktop.
6

We define a distributed systems as one in which hardware or
software components located at networked computers
communicate and coordinate their actions only by passing
messages.
Education System in Current Scenario (Online class)
8

Scalability is the property of a system to handle a growing
amount of work by adding resources to the system.
A system is described as scalable if it will remain effective
when there is a significant increase in the number of resources
and the number of users.
11

Taxonomy- from Greek taxis, meaning arrangement or
division
Michael J. Flynn (born May 20, 1934) is an American
professor at Stanford University.
Flynn proposed the Flynn’s taxonomy, a method of
classifying digital computer architectures, in 1966.
23

Computer architecture is a set of rules and methods that
describe the functionality, organization and
implementation of computer.
Computer architecture is concerned with balancing the
performance, efficiency, cost and reliability of a computer
system.
24

Instruction stream: The sequence of instructions from
memory to the control unit.
Data stream: The sequence of data from memory to
control unit.
27

A sequential computer which exploits no parallelism in
either the instruction or data streams.
Single control unit (CU) fetches single instruction stream
(IS) from memory.
The CU then generates appropriate control signals to
direct single processing element (PE) to operate on single
data stream (DS) i.e., one operation at a time.
29

Advantages
It requires less power.
There is no issue of complex communication protocol between
multiple cores.
Disadvantages
The speed of SISD architecture is limited just like single-core
processors.
It is not suitable for larger applications
30

A single instruction operates on multiple different data
streams.
Instructions can be executed sequentially, such as by
pipelining, or in parallel by multiple functional units.
Throughput of the system can be increased by increasing the
number of cores of the processor.
Processing speed is higher than SISD architecture.
There is complex communication between number of cores of
processor.
31

Multiple instructions operate on one data stream.
Systems with MISD stream have number of processing units
performing different operations by executing instructions on
the same data set.
32

Multiple autonomous processors simultaneously executing
different instructions on different data.
MIMD architecture include multi-core superscalar processors,
and distributed systems, using either one shared memory space
or a distributed memory space.
33

Process:- A program in execution is called process
Thread:- is the segment of a process, means a process can have
multiple threads and these multiple threads are contained
within a process.
35

39
In computer architecture, multithreading is the ability of a single CPU to
provide multiple threads of execution concurrently, supported by the OS.
Multithreading aims to increase utilization of a single core by using
thread level parallelism.
Multithreading allow for multiple requests to be satisfied simultaneously,
without having to service requests sequentially.

40
1. Responsiveness- If one thread completes its execution, then its output can
be immediately returned.
2. Faster Context switch- context switch time between threads is lower
compared to process context switch. Process context switching requires
more overhead from the CPU.
3. Effective utilization of multiprocessor system- If we have multiple threads
in a single process, then we can schedule multiple threads on multiple
processor. This will make process execution faster.
4. Resource sharing- Resources like code, data and files can be shared among
all threads within a process. Stack and register can’t be shared.

41
5. Communication- Communication between multiple threads is easier, as
compared to processes.
6. Enhanced throughput of the system- As each thread’s function is
considered as one job, then the number of jobs completed per unit of time
is increased, thus increasing the throughput of the system.

43
 Shared memory is memory that simultaneously accessed by multiple
programs with an intent to provide communication among them or avoid
redundant copies.
 Shared memory is an efficient means of passing data between programs.

44
 In computer hardware, shared memory refers to a (typically large) block
of Random Access Memory (RAM) that can be accessed by several
different CPUs in a multiprocessor computer system.

45
 In shared memory systems following different approaches may followed:-
 Uniformly Memory Access (UMA):- all the processors share the
physical memory uniformly (example-access time is same etc.)
 Non-uniform Memory Access (NUMA):-memory access time depends
on the memory location relative to a processor.
 Cache-only Memory Architecture (COMA):- the local memories for the
processors at each node is used as cache instead of as actual main
memory.

 Advantage:- A shared memory system is relatively easy to program since all
processors share a single view of data and the communication between
processors can be as fast as memory accesses to a same location.
 Challenge:- The issue with shared memory system is that many CPUs need
fast access to memory and will likely cache memory, which has two
complications.
o Access Time degradation- When several processors try to access the same
memory location, it causes contention.
o Lack of data coherence- whenever one cache is updated with information
that may be used by other processors, the change needs to be reflected to
the other processors otherwise the different processors will be working
with incoherent data.
47

 At software level, shared memory is either:
o A method of inter-process communication (IPC) i.e. a way of
exchanging data between programs running at the same time.
 Since both processes can access the shared memory area like
regular working memory, this is a very fast way of
communication.
 On the other hand, it is less scalable, and care must be taken to
avoid issues if processes sharing memory are running on separate
CPUs and the underlying architecture is not cache coherent.
48

Methodical Design of Parallel and Distributed
Algorithms
49

The model of parallel algorithms are developed by considering
a strategy for dividing the data and processing methods and
also applying a suitable strategy to reduce interactions.
Data Parallel Models
Task graph model
Work pool model
Master slave model
Producer consumer/pipeline model
Hybrid model
50

In data parallel model, tasks are assigned to processes and each task
performs similar types of operations on different data.
Data parallelism is a consequence of single operations that is being
applied on multiple data items.
Data-parallel model can be applied on shared-address spaces and message-
passing paradigms.
51

In the task graph model, parallelism is expressed by a task graph. A task
graph can be either trivial or nontrivial.
In this model, the correlation among the tasks are utilized to promote
locality or to minimize interaction costs.
This model is enforced to solve problems in which the quantity of data
associated with the tasks is huge compared to the number of computation
associated with them.
The tasks are assigned to help improve the cost of data movement among
the tasks.
Task has dependence with its predecessor and antecedent task
53

After completion of one task, the output is transferred to its dependent task.
Dependent task executes only when its antecedent tasks finishes execution.
54

 In work pool model, tasks are dynamically assigned to the processes
for balancing the load. Therefore, any process may potentially
execute any task.
 This model is used when the quantity of data associated with tasks is
comparatively smaller than the computation associated with the
tasks.
 There is no desired pre-assigning of tasks onto the processes.
 Pointers to the tasks are saved in a physically shared list, in a priority
queue, or in a hash table or tree, or they could be saved in a
physically distributed data structure.
55

 In the master-slave model, one or more master processes generate
task and allocate it to slave processes.The tasks may be allocated
beforehand if −
 the master can estimate the volume of the tasks, or
 a random assigning can do a satisfactory job of balancing load, or
 This model is generally equally suitable to shared-address-
space or message-passing paradigms, since the interaction is
naturally two ways.
57

 Care should be taken to assure that the master does not become a
congestion point. It may happen if the tasks are too small or the
workers are comparatively fast.
 The tasks should be selected in a way that the cost of performing a
task dominates the cost of communication and the cost of
synchronization.
59

 It is also known as the producer-consumer model. Here a set of
data is passed on through a series of processes, each of which
performs some task on it.
 Here, the arrival of new data generates the execution of a new task
by a process in the queue.
 This model is a chain of producers and consumers. Each process in
the queue can be considered as a consumer of a sequence of data
items for the process preceding it in the queue and as a producer of
data for the process following it in the queue.
60

 A hybrid algorithm model is required when more than one model
may be needed to solve a problem.
 A hybrid model may be composed of either multiple models applied
hierarchically or multiple models applied sequentially to different
phases of a parallel algorithm.
62

 Parallel Random Access Machines (PRAM) is a model, which
is considered for most of the parallel algorithms. Here, multiple
processors are attached to a single block of memory. A PRAM
model contains −
 A set of similar type of processors.
 All the processors share a common memory unit. Processors can
communicate among themselves through the shared memory only.
 A memory access unit (MAU) connects the processors with the
single shared memory.
63

 Here, n number of processors can perform independent operations
on n number of data in a particular unit of time.This may result in
simultaneous access of same memory location by different
processors.
 To solve this problem, the following constraints have been enforced
on PRAM model −
 Exclusive Read ExclusiveWrite (EREW) − Here no two
processors are allowed to read from or write to the same memory
location at the same time.
 Exclusive Read ConcurrentWrite (ERCW) − Here no two
processors are allowed to read from the same memory location at
the same time, but are allowed to write to the same memory
65

 Concurrent Read ExclusiveWrite (CREW) − Here all the
processors are allowed to read from the same memory location at
the same time, but are not allowed to write to the same memory
 Concurrent Read ConcurrentWrite (CRCW) − All the
processors are allowed to read from or write to the same memory
66

Cloud Platforms
68
 Cloud platforms is the delivery of different services through
the internet.
 These resources include tools and applications like:
 Data storage
 Servers
 Databases
 Networking
 Software
 Cloud platforms are popular options that saves cost,
increased productivity, speed and efficiency, performance and
security.

Cloud Platforms examples
69
 Amazon Cloud computing (AWS)
 Google Cloud Platform
 Azure Cloud Computing (Microsoft Azure)
 Hadoop
 Force.com and Salesforce.com

Amazon Cloud computing (AWS)
70

71
 AWS Compute Services
 Elastic Compute Cloud (EC2)
 LightSail
 Elastic Beanstalk
 EKS (Elastic Container Service for Kubernetes)
 AWS Lambda
 AWS Migration Services
 DMS (Database Migration Services)
 SMS (Server Migration Services)
 Snowball

72
 AWS Storage Services
 Amazon Glacier
 Amazon Elastic Block Store (EBS)
 AWS Storage Gateway
 AWS Database Services
 Amazon RDS
 Amazon DynamoDB
 Neptune
 Amazon RedShift

73
 AWS Security Services
 IAM
 Inspector
 Certificate Manager
 WAF
 Cloud Directory
 KMS
 Shield

Microsoft Azure
74
 AWS Analytics Services
 Athena
 CloudSearch
 ElasticSearch
 QuickSight
 Data Pipeline
 Management Services
 CloudWatch
 CloudFormation
 ServiceCatalog
 AWSAutoscaling

Azure Service-Compute
77
 Cloud Service
 Service Fabric
 Functions

Azure Service-Networking
78
 Azure CDN
 Express Route
 Virtual Network
 Azure DNS

Azure Service-Storage
79
 Disk Storage
 Blob Storage
 File Storage
 Queue Storage

Azure Applications
80
 Application development
 Testing
 Application hosting
 CreatingVirtual Machine
 Virtual hard drives

Azure Advantage
81
 Storage Efficiency
 Cost effective
 Stability
 Security

Google Cloud Platform (GCP)
82
 GCP is offered by Google
 It is a suite of cloud computing services that runs on the same
infrastructure that google uses internally for its end users
products
 provides a set of management tools it provides a series of
cloud services : computing, data storage, data analytics,
machine learning
 GCP is IaaS of google platform.

Features of GCP
83
 Virtual Machines
 Local SSD
 Persistent Disks
 GPUAccelerators
 Global Load Balancing

Module 2.pdf

More Related Content

Similar to Module 2.pdf (20)

Recently uploaded (20)

Module 2.pdf