Chapter 4 – Cloud Computing

Applications and Paradigms

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 1
Contents
 Challenges for cloud computing.
 Existing cloud applications and new opportunities.
 Architectural styles for cloud applications.
 Workflows - coordination of multiple activities.
 Coordination based on a state machine model.
 The MapReduce programming model.
 A case study: the GrepTheWeb application.
 Clouds for science and engineering.
 High performance computing on a cloud.
 Legacy applications on a cloud.
 Social computing, digital content, and cloud computing.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 2
Cloud applications
 Cloud computing is very attractive to the users:
 Economic reasons.
 low infrastructure investment.
 low cost - customers are only billed for resources used.
 Convenience and performance.
 application developers enjoy the advantages of a just-in-time
infrastructure; they are free to design an application without being
concerned with the system where the application will run.
 the execution time of compute-intensive and data-intensive
applications can, potentially, be reduced through parallelization.
If an application can partition the workload in n segments and
spawn n instances of itself, then the execution time could be
reduced by a factor close to n.
 Cloud computing is also beneficial for the providers of computing
cycles - it typically leads to a higher level of resource utilization.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 3
Cloud applications (cont’d)
 Ideal applications for cloud computing:
 Web services.
 Database services.
 Transaction-based service. The resource requirements of transaction-
oriented services benefit from an elastic environment where resources
are available when needed and where one pays only for the resources
it consumes.
 Applications unlikely to perform well on a cloud:
 Applications with a complex workflow and multiple dependencies, as
is often the case in high-performance computing.
 Applications which require intensive communication among concurrent
instances.
 When the workload cannot be arbitrarily partitioned.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 4
Challenges for cloud application development
 Performance isolation - nearly impossible to reach in a real system,
especially when the system is heavily loaded.

 Reliability - major concern; server failures expected when a large

number of servers cooperate for the computations.

 Cloud infrastructure exhibits latency and bandwidth fluctuations

which affect the application performance.

 Performance considerations limit the amount of data logging; the

ability to identify the source of unexpected results and errors is
helped by frequent logging.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 5
Existing and new application opportunities

 Three broad categories of existing applications:

 Processing pipelines.
 Batch processing systems.
 Web applications.
 Potentially new applications
 Batch processing for decision support systems and business
analytics.
 Mobile interactive applications which process large volumes of
data from different types of sensors.
 Science and engineering could greatly benefit from cloud
computing as many applications in these areas are compute-
intensive and data-intensive.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 6
Processing pipelines
 Indexing large datasets created by web crawler engines.
 Data mining - searching large collections of records to locate items
of interests.
 Image processing .
 Image conversion, e.g., enlarge an image or create thumbnails.
 Compress or encrypt images.
 Video transcoding from one video format to another, e.g., from AVI
to MPEG.
 Document processing.
 Convert large collections of documents from one format to
another, e.g., from Word to PDF.
 Encrypt documents.
 Use Optical Character Recognition to produce digital images of
documents.
Cloud Computing: Theory and Practice.
Dan C. Marinescu Chapter 4 7
Batch processing applications
 Generation of daily, weekly, monthly, and annual activity reports for
retail, manufacturing, other economical sectors.

 Processing, aggregation, and summaries of daily transactions for

financial institutions, insurance companies, and healthcare
organizations.

 Processing billing and payroll records.

 Management of the software development, e.g., nightly updates of

software repositories.

 Automatic testing and verification of software and hardware

systems.
Cloud Computing: Theory and Practice.
Dan C. Marinescu Chapter 4 8
Web access

 Sites for online commerce.

 Sites with a periodic or temporary presence.

 Conferences or other events.
 Active during a particular season (e.g., the Holidays Season) or
income tax reporting.

 Sites for promotional activities.

 Sites that ``sleep'' during the night and auto-scale during the
day.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 9
Architectural styles for cloud applications
 Based on the client-server paradigm.
 Stateless servers - view a client request as an independent
transaction and respond to it; the client is not required to first
establish a connection to the server.
 Often clients and servers communicate using Remote Procedure
Calls (RPCs).
 Simple Object Access Protocol (SOAP) - application protocol for
web applications; message format based on the XML. Uses TCP
or UDP transport protocols.
 Representational State Transfer (REST) - software architecture
for distributed hypermedia systems. Supports client
communication with stateless servers, it is platform independent,
language independent, supports data caching, and can be used in
the presence of firewalls.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 10
Workflows
 Process description - structure describing the tasks to be
executed and the order of their execution. Resembles a flowchart.

 Case - an instance of a process description.

 State of a case at time t - defined in terms of tasks already

completed at that time.

 Events - cause transitions between states.

 The life cycle of a workflow - creation, definition, verification, and

enactment; similar to the life cycle of a traditional program
(creation, compilation, and execution).
Cloud Computing: Theory and Practice.
Dan C. Marinescu Chapter 4 11
 Dynamic Workflows Static Workflows Static Programs Dynamic Programs
Workflow
Programming
Description Component
Component Language
Language Libraries
Database
User User
Planning Automatic
Engine Programming

Workflow Computer
Description Program

Verification
Engine Compiler

Workflow Object
Workflow Program
Description Code
Database Libraries

Case Activation Record Data

Processor
Enactment
Running
Engine Run-Time Program
Unanticipated Exception the Process
Handling Modification Requests

(a) Workflow (b) Program

Cloud Computing: Theory and Practice.
Dan C. Marinescu Chapter 4 12
Safety and liveness

 Desirable properties of workflows.

 Safety  nothing “bad” ever happens.

 Liveness  something “good” will eventually happen.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 13
 Cloud Computing: Theory and Practice.
Dan C. Marinescu Chapter 4 14
 Basic workflow patterns
 Workflow patterns - the temporal relationship among the tasks of a process
 Sequence - several tasks have to be scheduled one after the completion of
the other.
 AND split - both tasks B and C are activated when task A terminates.
 Synchronization - task C can only start after tasks A and B terminate.
 XOR split - after completion of task A, either B or C can be activated.
 XOR merge - task C is enabled when either A or B terminate.
 OR split - after completion of task A one could activate either B, C, or both.
 Multiple Merge - once task A terminates, B and C execute concurrently;
when the first of them, say B, terminates, then D is activated; then, when C
terminates, D is activated again.
 Discriminator – wait for a number of incoming branches to complete before
activating the subsequent activity; then wait for the remaining branches to
finish without taking any action until all of them have terminated. Next,
resets itself.
Cloud Computing: Theory and Practice.
Dan C. Marinescu Chapter 4 15
 Basic workflow patterns (cont’d)
 N out of M join - barrier synchronization. Assuming that M tasks
run concurrently, N (N<M) of them have to reach the barrier before
the next task is enabled. In our example, any two out of the three
tasks A, B, and C have to finish before E is enabled.
 Deferred Choice - similar to the XOR split but the choice is not
made explicitly; the run-time environment decides what branch to
take.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 16
 B A

A B C A AND AND C
C B

a b c

B A B

A XOR XOR C A OR

C B C

d e f

B B

A AND XOR D A AND DIS D

C C

g h

B B

A XOR X
A AND C 2/3 E
C

i j

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 17
Coordination - ZooKeeper
 Cloud elasticity  distribute computations and data across multiple
systems; coordination among these systems is a critical function in a
distributed environment.
 ZooKeeper
 Distributed coordination service for large-scale distributed systems.
 High throughput and low latency service.
 Implements a version of the Paxos consensus algorithm.
 Open-source software written in Java with bindings for Java and C.
 The servers in the pack communicate and elect a leader.
 A database is replicated on each server; consistency of the replicas is
maintained.
 A client connect to a single server, synchronizes its clock with the
server, and sends requests, receives responses and watch events
through a TCP connection.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 18
 Server Server Server Server Server

Client Client Client Client Client Client Client Client

(a)

Follower
Replicated
database Follower
Write
processor Leader
Follower

Follower
Atomic broadcast Follower

WRITE READ WRITE

(b) (c)
Cloud Computing: Theory and Practice.
Dan C. Marinescu Chapter 4 19
Zookeeper communication
 Messaging layer  responsible for the election of a new leader
when the current leader fails.

 Messaging protocols use:

 Packets - sequence of bytes sent through a FIFO channel.
 Proposals - units of agreement.
 Messages - sequence of bytes atomically broadcast to all
servers.
 A message is included into a proposal and it is agreed upon
before it is delivered.
 Proposals are agreed upon by exchanging packets with a
quorum of servers, as required by the Paxos algorithm.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 20
Zookeeper communication (cont’d)

 Messaging layer guarantees:

 Reliable delivery: if a message m is delivered to one server, it will

be eventually delivered to all servers.

 Total order: if message m is delivered before message n to one

server, it will be delivered before n to all servers.

 Causal order: if message n is sent after m has been delivered by

the sender of n, then m must be ordered before n.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 21
Shared hierarchical namespace similar to a
file system; znodes instead of inodes

/a /b /c

/a/1 /a/2 /b/1 /c/1 /c/2

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 22
ZooKeeper service guarantees
 Atomicity - a transaction either completes or fails.

 Sequential consistency of updates - updates are applied strictly

in the order they are received.

 Single system image for the clients - a client receives the same
response regardless of the server it connects to.

 Persistence of updates - once applied, an update persists until

it is overwritten by a client.

 Reliability - the system is guaranteed to function correctly as

long as the majority of servers function correctly.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 23
Zookeeper API
 The API is simple - consists of seven operations:

 Create - add a node at a given location on the tree.

 Delete - delete a node.

 Get data - read data from a node.

 Set data - write data to a node.

 Get children - retrieve a list of the children of the node.

 Synch - wait for the data to propagate.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 24
Elasticity and load distribution
 Elasticity  ability to use as many servers as necessary to optimally
respond to cost and timing constraints of an application.
 How to divide the load
 Transaction processing systems  a front-end distributes the incoming
transactions to a number of back-end systems. As the workload
increases new back-end systems are added to the pool.
 For data-intensive batch applications two types of divisible workloads are
possible:
 modularly divisible  the workload partitioning is defined a priori.

 arbitrarily divisible  the workload can be partitioned into an

arbitrarily large number of smaller workloads of equal, or very close
size.
 Many applications in physics, biology, and other areas of
computational science and engineering obey the arbitrarily divisible
load sharing model.
Cloud Computing: Theory and Practice.
Dan C. Marinescu Chapter 4 25
MapReduce philosophy
1. An application starts a master instance, M worker instances for the
Map phase and later R worker instances for the Reduce phase.
2. The master instance partitions the input data in M segments.
3. Each map instance reads its input data segment and processes
the data.
4. The results of the processing are stored on the local disks of the
servers where the map instances run.
5. When all map instances have finished processing their data, the R
reduce instances read the results of the first phase and merge the
partial results.
6. The final results are written by the reduce instances to a shared
storage server.
7. The master instance monitors the reduce instances and when all of
them report task completion the application is terminated.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 26
 Application

Master instance

1 1 7

Map
Segment 1
instance 1 Local disk
Reduce
Segment 12 Map instance 1
Segment instance 2 Local disk Shared
Map Reduce storage
Segment 3 instance 3 Local disk instance 2

Shared
storage
Reduce
3 4 5 instance R 6
Map
Segment M instance M Local disk

Input data Map phase Reduce phase

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 27
Case study: GrepTheWeb
 The application illustrates the means to
 create an on-demand infrastructure.
 run it on a massively distributed system in a manner that allows
it to run in parallel and scale up and down, based on the number
of users and the problem size.
 GrepTheWeb
 Performs a search of a very large set of records to identify
records that satisfy a regular expression.
 It is analogous to the Unix grep command.
 The source is a collection of document URLs produced by the
Alexa Web Search, a software system that crawls the web every
night.
 Uses message passing to trigger the activities of multiple
controller threads which launch the application, initiate
processing, shutdown the system, and create billing records.
Cloud Computing: Theory and Practice.
Dan C. Marinescu Chapter 4 28
 Input records
(a) The simplified workflow Regular
expression SQS

showing the inputs: Controller

- the regular expression. Output
EC2
- the input records generated Status Simple
DB Cluster S3

by the web crawler. (a)

- the user commands to report
the current status and to Billing
queue
terminate the processing. Launch
queue
Monitor
queue

Shutdown
(b) The detailed workflow. queue
Billing
service
The system is based on
message passing between Controller

several queues; four Launch Monitor Shutdown Billing

controller controller controller controller
controller threads
periodically poll their
Put file
associated input queues, Status
Output

DB
retrieve messages, and HDHS Input
Get file
carry out the required Hadoop Cluster on
Amazon SimpleDB Amazon S3
actions Amazon SE2

(b)
Cloud Computing: Theory and Practice.
Dan C. Marinescu Chapter 4 29
Clouds for science and engineering

 The generic problems in virtually all areas of science are:

 Collection of experimental data.
 Management of very large volumes of data.
 Building and execution of models.
 Integration of data and literature.
 Documentation of the experiments.
 Sharing the data with others; data preservation for a long periods of
time.

 All these activities require “big” data storage and systems capable
to deliver abundant computing cycles.
Computing clouds are able to provide such resources and support
collaborative environments.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 30
Online data discovery
 Phases of data discovery in large scientific data sets:
 recognition of the information problem.
 generation of search queries using one or more search engines.
 evaluation of the search results.
 evaluation of the web documents.
 comparing information from different sources.
 Large scientific data sets:
 biomedical and genomic data from the National Center for
Biotechnology Information (NCBI).
 astrophysics data from NASA.
 atmospheric data from the National Oceanic and Atmospheric
Administration (NOAA) and the National Center for Atmospheric
Research (NCAR).

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 31
High performance computing on a cloud
 Comparative benchmark of EC2 and three supercomputers at the
National Energy Research Scientific Computing Center (NERSC) at
Lawrence Berkeley National Laboratory. NERSC has some 3,000
researchers and involves 400 projects based on some 600 codes.
 Conclusion – communication-intensive applications are affected by
the increased latency and lower bandwidth of the cloud. The low
latency and high bandwidth of the interconnection network of a
supercomputer cannot be matched by a cloud.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 32
Legacy applications on the cloud
 Is it feasible to run legacy applications on a cloud?
 Cirrus - a general platform for executing legacy Windows
applications on the cloud. A Cirrus job - a prologue, commands,
and parameters. The prologue sets up the running environment; the
commands are sequences of shell scripts including Azure-storage-
related commands to transfer data between Azure blob storage and
the instance.
 BLAST - a biology code which finds regions of local similarity
between sequences; it compares nucleotide or protein sequences
to sequence databases and calculates the statistical significance of
matches; used to infer functional and evolutionary relationships
between sequences and identify members of gene families.
 AzureBLAST - a version of BLAST running on the Azure platform.

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 33
 Cirrus

Web role Job manager role

Scaling Worker
Web engine
portal
Job Job Worker
registration scheduler Parametric
engine
Web
Dispatch queue Worker
service
Sampling
filter Worker

Worker
Azure table
Azure blob
Worker

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 34
Execution of loosely-coupled workloads using the
Azure platform
Client

BigJob Manager

start VM start replicas query state

Queues
Blob

Portal
Service
post results
Mahagement
API

query

Worker Role Worker Role

BigJob Agent BigJob Agent
task k task task task n
task 1 task 1
k+1 k+2

Cloud Computing: Theory and Practice.

Dan C. Marinescu Chapter 4 35
Social computing and digital content
 Networks allowing researchers to share data and provide a virtual
environment supporting remote execution of workflows are domain
specific:
 MyExperiment for biology.
 nanoHub for nanoscience.
 Volunteer computing - a large population of users donate resources
such as CPU cycles and storage space for a specific project:
 Mersenne Prime Search
 SETI@Home,
 Folding@home,
 Storage@Home
 PlanetLab
 Berkeley Open Infrastructure for Network Computing (BOINC) 
middleware for a distributed infrastructure suitable for different
applications.
Cloud Computing: Theory and Practice.
Dan C. Marinescu Chapter 4 36