Unit 5: Cloud Applications

Cloud Applications

• Cloud computing has gained huge popularity in industry due to its ability to host applications whose
services can be delivered to consumers rapidly at minimal cost.

• This chapter discusses various application case studies detailing their architecture and how they
leveraged various Cloud technologies.

• Applications from a range of domains from scientific to engineering, gaming, to social networking are
considered.

Scientific Applications

• Scientific applications are a sector that is increasingly using Cloud computing systems and technologies.

• The immediate benefit seen by researchers and academics is the potentially infinite availability of
computing resources and storage at sustainable prices if compared to a complete in-house deployment.

. Cloud computing systems meet the needs of different types of applications in the scientific domain: HPC
(High Performance Computing) applications, HTC (High Throughput Computing) applications, and data-
intensive applications.

The opportunity for using Cloud resources is even more appealing since minimal changes need to be done
to existing applications in order to leverage Cloud resources.

Healthcare: ECG Analysis in the Cloud

Healthcare is a domain where computer technology has found several and diverse applications: from
supporting the business functions to assisting scientists in developing solutions to cure diseases.

• An important application is the use of Cloud technologies for supporting doctors in providing more
effective diagnostic processes.

In particular, we discuss electrocardiogram (ECG) data analysis on the Cloud.

• The capillary development of Internet connectivity and its accessibility from any device at any time has
made Cloud technologies an attractive option for developing health-monitoring systems. •
Electrocardiogram (ECG) data analysis and monitoring constitutes a case study that naturally fits in this
scenario. ECG is the electrical manifestation of the contractive activity of the heart's myocardium.

• This activity produces a specific waveform that is repeated overtime and that represents the heartbeat.
The analysis of the shape of the waveform is used to identify arrhythmias and it is the most common way
for detecting heart diseases.
• Cloud computing technologies allow the remote monitoring of a patient's heartbeat data, its analysis in
minimum time, and the notification of first-aid personnel and doctors should this data reveal potentially
dangerous conditions,

• This way a patient at risk can be constantly monitored without going to hospital for ECG analysis. At
the same time, doctors and first-aid personnel can instantly be notified with cases that require their
attention.

. Even though remote ECG monitoring does not necessanly require Cloud technologies, Cloud computing
introduces opportunities that would be otherwise hardly achievable,

• The first advantage is the elasticity of the Cloud infrastructure that can grow and shrink according to the
requests served. As a result, doctors and hospitals do not have to invest in large computing infrastructures
designed after capacity planning. thus making a more effective use of budgets. The second advantage is
ubiquity

• Cloud computing technologies have now become easily accessible and promise to deliver systems with
minimum or no downtime Computing systems hosted in the Cloud are accessible from any Internet
device through simple interfaces (such as SOAP and REST based web services)

. This makes not only these systems ubiquitous but they can also be easily integrated with other systems
maintained in the hospital's premises Lasty cost savings constitute another reason.

*Cloud services are priced on a pay-per-use basis and with volume prices in case of large numbers of
service requests. These two models provide a set of flexible options that can be used to price the service,
thus actually charging costs based on effective use rather than capital costs

Biology: Protein Structure Prediction

• Applications in biology often require high computing capabilities and often operate on large datasets
that cause extensive VO operations

• Because of these requirements, they have aften made extensive use of supercomputing and cluster
computing infrastructures. Similar capabilities can be leveraged on-demand by using Cloud computing
technologies in a more dynamic fashion thus opening new opportunities for bioinformatics applications
Protein structure prediction is a computationally intensive task fundamental for different types of research
in the life sciences.

• Among these is the design of new drugs for the treatment of diseases. The geometrical structure of a
protein cannot be directly inferred from the sequence of genes that compose its structure, but it is the
result of complex computations aimed at identifying the structure that minimizes the required energy.

This task requires the investigation of a space with a massive number of states, and consequently creating
a large number of computations for each of these states. The computational power required for protein
structure prediction can now be acquired on demand, without owning a cluster or doing all the
bureaucracy for getting access to parallel and distributed computing facilities. Cloud computing grants the
access to such capacity on a pay-per-use basis

Biology: Gene Expression Data Analysis for Cancer Diagnosis

Gene expression profing is the measurement of the expression levels of thousands of genes at orice. It is
used to understand the biological processes that are triggered by the treatment at a celular level.

Together with protein structure prediction, this activity is a fundamental component of drug design since
it allows scientists to identify the effects of a specific treatment • Another important application of gene
expression profiling is cancer diagnosis and treatment Cancer is a disease characterized by uncontroled
cell growth and proliferation.

• This behavior occurs because of genes regulating the cel growth mutate. This means that all the
cancerous cells contain mutated genes

• In this context gene expression profiling is utilized to provide a more accurate classification of tumors.
• The classification of gene expression data samples into destinct classes is a challenging task. The
dimensionality of typical gme expression data sets ranges from several thousands to over ten thousands
genes. However, only small sample sizes are typically available for analysis,

Business and Consumer Applications

• The business and consumer sector is the one that probably benefits the most from Cloud computing
technologies.

• On the one hand the opportunity of transforming capital cost into operational costs makes Clouds an
attractive option for all enterprises that are IT centric.

. On the other hand, the sense of ubiquity that Cloud offers for accessing data and services makes it
interesting for end users as well.

Moreover, the elastic nature of Cloud technologies does not require huge upfront investments, thus
allowing new ideas to be quickly translated into products and services that can comfortably grow with the
demand.

• The combination of all these elements has made Cloud computing the preferred technology for a wide
range of applications: from CRM and ERP systems to productivity and social networking applications.

CRM and ERP

• Customer Relationship Management (CRM) and Enterprise Resource Planning (ERP) applications are
market segments that are flourishing in the Cloud, with CRM applications being more mature than ERP
implementations.

• Cloud CRM applications constitute a great opportunity for small enterprises and start-ups to have a fully
functional CRM software without large upfront costs and by paying subscriptions.

• Moreover, customer relationship management is not an activity that requires specific needs and it can be
easily moved to the Cloud. Such a characteristic, together with the possibility of having access to your
business and customer data from everywhere and any device, has fostered the spread of Cloud CRM
applications. ERP solutions on the Cloud are less mature and have to compete with well established in-
house solutions.

• ERP systems integrate several aspects of an enterprise: finance and accounting, human resources,
manufacturing, supply chain management, project management, and customer relationship management.

SalesForce.com

• Salesforce.com is probably the most popular and developed CRM solutions available today. As of today
more than 100 thousands customers have chosen Safesforce.com to implement their CRM solutions.

• The application provides customizable CRM solutions that can be integrated with additional features
developed by third parties. Salesforce.com is based on the Force.com Cloud development platform.

• This represents the scalable and high-performance middleware executing all the operations of all
Salesforce.com applications.

Social Networking

• Social networking applications have considerably grown in the last years to become the most active sites
on the web. In order to sustain their traffic and to serve millions of users. seamlessly,

• services like Twitter or Facebook, have leveraged Cloud computing technologies.

• The possibility of continuously adding capacity while systems are running is the most attractive feature
for social networks, which constantly increase their user base.

Facebook

• Facebook is probably the most evident and interesting environment in social networking. It became one
of the largest web sites in the world with more than 800 million users. In order to sustain this incredible
growth it has been fundamental to be capable of continuously adding capacity, developing new scalable
technologies and software systems while keeping a high performance for a smooth user experience.

. Currently, the social network is backed by two data centers that have been built and optimized to reduce
costs and impact on the environment. On top of this highly efficient infrastructure built and designed out
of inexpensive hardware, a completely customized stack of open source technologies opportunely
modified and refined constitutes the backend of largest social network.
• Taken all together, these technologies constitute a powerful platform for developing Cloud applications.

• This platform primarily supports Facebook itself and offers APIs to integrate third party applications
with Facebook's core infrastructure to deliver additional services such as social games and quizzes created
by others.

• The reference stack serving Facebook is based on LAMP (Linux, Apache, MySQL, and PHP). This
collection of technologies is accompanied by a collection of other services developed in-house.

• These services are developed in a variety of languages and implement specific functionalities such as
search, new feeds, notifications, and others.
• While serving page requests, the social graph of the user is composed. The social graph identifies
collection of interlinked information that is of relevance for a given user.

• Most of the user data is served by querying a distributed cluster of MySQL instances, which mostly
contain key-value pairs. This data is then cached for faster retrieval.
•The rest of the relevant information is then composed together by using the services mentioned before.

• These services are located closer to the data and developed in languages that provide a better
performance than PHP.
• The development of services is facilitated by a set of tools internally developed. One of the core
elements is Thrift.

• This is a collection of abstractions (and language bindings) that allow cross-language development.
Thrift allows services developed in different languages to communicate and exchange data. Bindings for
Thrift in different languages take care of data serialization and deserialization, communication, and client
and server boilerplate code.

• This simplifies the work of the developers that can quickly prototype services and leverage existing one.
Other relevant services and tools are Scribe, which aggregates streaming log feeds, and applications for
alerting and monitoring.

Media Applications

• Media applications are a niche that has taken a considerable advantage from leveraging Cloud
computing technologies.

• In particular, video processing operations, such as encoding, transcoding, composition, and rendering,
candidates for a Cloud-based environment. are good

• These are computationally intensive tasks that can be easily offloaded to Cloud computing
infrastructures.
Animoto

• Animoto is perhaps the most popular example of media applications on the Cloud. The website provides
users with a very straightforward interface for quickly creating videos out of images, music, and video
fragments submitted by users.

• Users select a specific theme for the video, upload the photos and videos and order them in the sequence
they want to appear, select the song for the music, and render the video.

• The process is executed in the background and the user is notified via e-mail once the video is rendered.

• The core value of Animoto is the ability to quickly create videos with stunning effects without the user
intervention. • A proprietary Al engine that selects the animation and transition effects according to
pictures and music drives the rendering operation.

• Users only have to define the storyboard by organizing pictures and videos into the desired sequence. If
not, the video can be rendered again and the engine will select a different composition, thus producing a
different outcome every time. The service allows creating 30 seconds videos for free.

• By paying a monthly or a yearly subscription it is possible to produce videos of any length and to
choose among a wider range of templates.

Maya Rendering with Aneka

• Interesting applications of media processing are found in the engineering disciplines and the movie
production industry. Operations such as rendering of models are now an integral part of the design
workflow, which has become computationally demanding.

• The visualization of mechanical models is not only used at the end of the design process, but it is
iteratively used to improve the design.

It is then fundamental to perform such task as fast as possible. Cloud computing provides engineers with
the necessary computing power to make this happen.

Video Encoding on the Cloud : Encoding.com

• Video encoding and transcoding are operations that can take a great benefit from using Cloud
technologies: they are computationally intensive and potentially require considerable amount of storage.

• Moreover, with the continuous improvement of mobile devices as well as the diffusion of Internet,
requests for video content have significantly increased.

• The variety of devices with video playback capabilities has led to an explosion of video formats through
which a video can be delivered.
. Software and hardware for video encoding and transcoding often have prohibitive costs or are not
flexible enough to support conversion from any format to any format.

• Cloud technologies present an opportunity for turning these tedious and often demanding tasks into
services that can be easily integrated into different workflows or made available to everyone according to
their needs.
• Encoding.com is software solution that offers video transcoding services on demand and leverage Cloud
technology to provide both the horse-power required for video conversion and the storage for staging
videos. • The service integrates both with Amazon Web Services technologies (EC2,S3, and CloudFront)
and Rackspace (Cloud Servers, Cloud Files, and Limelight CDN access).

• Users can access the services through a variety of interfaces: Encoding.com website, web service XML
APls, desktop applications, and watched folders.

• In order to use the service users have to specify the location of the video to transcode, the destination
format, and the target location of the video. Encoding.com also offers other video editing operations such
the insertion of thumbnails, watermarks, or logos. Moreover, it also extends its capabilities to audio and
image conversion.

• The service provides different pricing options: monthly fee, pay-as-you-go (by batches), and special
prices for high volumes. Encoding.com has up to now more than 2000 customers and has already
processed more than 10 million videos.

Multiplayer Online Gaming

• Online multiplayer gaming attracts millions of gamers around the world that share a common experience
by playing together on a virtual environment that extends beyond the boundaries of a normal LAN.

• Online games support hundreds of players in the same session and this is made possible by the specific
architecture used to forward interactions that is based on game log processing.

Players update the game server hosting the game session and the server integrates all the updates into a
log that is made available to all the players through a TCP port.

• The client software used for the game connects to the log port and by reading the log updates the local
user interface with the actions of other players.
• Game log processing is also utilized to build statistics on players and rank them. These features
constitute the additional value of online gaming portals that attract more and more gamers.

• The processing of game logs is a potentially compute intensive operation that strongly depends on the
number of players online and the number of games monitored.

• Moreover, gaming portals are web applications and therefore might suffer from the spiky behavior of
users that can randomly generate large amount of volatile workloads that do not justify capacity planning.
• The use of Cloud computing technologies can provide the required elasticity for seamlessly processing
these workloads and scale as required when the number of users increases.
• A prototypal implementation of Cloud-based game log processing has been implemented by Titan Inc.
(now Xfire), a company based in California that extended its gaming portal to offload game log
processing to the Cloud by using Aneka.

• The prototype has utilized a private Cloud deployment that has allowed Titan Inc. to process
concurrently multiple logs and sustain a larger number of users.