COMPARISON OF OPEN-SOURCE PAAS

ARCHITECTURAL COMPONENTS
Mohan Krishna Varma Nandimandalam1 and Eunmi Choi2
1
Graduate School of Business IT, Kookmin University, Seoul, South Korea
[email protected]
2
Corresponding Author, School of Business IT,
Kookmin University, Seoul, Korea
[email protected]

ABSTRACT
Cloud computing is a widely used technology with three basic service models such as Software
as a Service (SaaS), Platform as a Service (PaaS), and Infrastructure as a Service (IaaS). This
paper focuses on the PaaS model. Open source PaaS model provides choice of cloud, developer
framework and application service. In this paper detailed study of four open PaaS packages
such as AppScale, Cloud Foundry, Cloudify, and OpenShift are explained with the considerable
architectural component aspects. We also explained some other PaaS packages like Stratos,
Stakato and mOSAIC briefly. In this paper we present the comparative study of major open
PaaS packages.

KEYWORDS
Cloud Computing, AppScale, Cloud Foundry, Cloudify, OpenShift, Stackato & Stratos

1. INTRODUCTION
Cloud computing is an emerging paradigm in which computers are networked to provide storage
and compute services using virtualization technology. Cloud computing must satisfy five
essential characteristics. They are on demand service, access network, resource pooling, elasticity
and measured services. To achieve these five essential characteristics, cloud computing provides
three kinds of service models: Software as a Service (SaaS), Platform as a Service (PaaS) [7] and
Infrastructure as a Service (IaaS) [8]. Cloud computing service models are shown in Figure 1.
CRM applications are widely used services in the SaaS. Application platform delivered as a
service is described as PaaS and it is used to deploy the user code. AppScale [2], Cloud Foundry,
Cloudify and OpenShift open-source environments can be used as PaaS. IaaS is used to build
their private infrastructure, which reduces the setup cost. IaaS can provide virtualized resources
such as computation, storage and communication. Eucalyptus [1], open stack and cloud stack
open-sources can be used to provide IaaS.

David C. Wyld et al. (Eds) : CCSEA, CLOUD, DKMP, SEA, SIPRO - 2016
pp. 53–62, 2016. © CS & IT-CSCP 2016 DOI : 10.5121/csit.2016.60206
54 Computer Science & Information Technology (CS & IT)

Figure 1. Cloud computing service models

This paper will focus on the PaaS service model. It is easy to deploy, run and scale application
using PaaS. Some of the PaaS have limited language and framework support. They do not deliver
key application services needed for cloud applications. They sometime restrict deployment to a
single cloud. Whereas open PaaS provides choice of cloud like private, public or hybrid, choice
of developer framework like spring, ruby, or java and application services like mongoDB,
MySQL, or PostgreSQL for running our applications. This paper deals with the architectural
components of major open PaaS packages like AppScale, Cloud Foundry, Cloudify and
OpenShift.

The paper is organized as follows. Section 2 introduce AppScale and its components, Cloud
Foundry architecture and component explanation given in Section 3, Cloudify open PaaS is
explained in Section 4, Section 5 deals with OpenShift, other open PaaS technologies are
introduced in Section 6, comparison of open-source PaaS technologies are given in Section 7 and
finally Section 8 concludes the paper.

2. APPSCALE

AppScale [3] is a scalable, distributed, and fault-tolerant cloud runtime system that executes over
cluster resources. It can be deployed on Xen [4], Kernel-based Virtual Machine (KVM), Amazon
EC2 or Eucalyptus. AppScale initial design utilizes the standard three-tier web deployment model
in the design. In the later design cycles more components are added to the AppScale. Table 1
shows the AppScale components, language used to design the component and their functionality.

Table 1. AppScale Components

Component Language Functionality

AppController Ruby Executes on every node and starts automatically when the
guest virtual machine boots
AppLoadBalancer Ruby on Rails Processes arriving requests from users and forwards them to
the application server
AppServer Python Running through a number of distant hosts to support
automated execution of applications
Database Master Python Offers persistent storage for applications, processes protocol
buffers from apps and makes requests on its behalf to read
and write data to the data store
Database Slave Python Facilitate distributed, scalable, and fault tolerant data
management
AppScale Tools Ruby Starts an AppScale system, deploys and tear down
applications, queries the state and performance of AppScale
deployment or application, and manipulates AppScale
configuration and state
 Computer Science & Information Technology (CS & IT) 55

3. CLOUD FOUNDRY
Cloud Foundry [10] is an open PaaS, which provides choice of clouds, developer frameworks and
application services. Cloud Foundry makes application development faster and easier. We can
build, test, deploy and scale applications with help of Cloud Foundry. It is an open-source project
available through a variety of private cloud distributions and public cloud instances. Cloud
Foundry started as a platform to deploy Java Spring applications on Amazon Web Services.
VMware acquired the Cloud Foundry and made it into an open-source, multi-language and multi-
framework PaaS. Cloud Foundry supports multiple languages and multiple runtimes such as Java,
Ruby, Scala, spring and Node.js. Cloud Foundry can run on anything like laptop, desktop, micro
cloud, private cloud or public cloud. So, it is called as open PaaS as shown in Figure 2. Cloud
Foundry has three dimensions to the platform: choice of frameworks, choice of application
services and the deployment choice. Cloud Foundry supports spring for Java, Rails and Sinatra
for Ruby, Node.js and JVM languages like Groovy, Grails and Scala. It also supports Microsoft
.NET Framework and became the first non-Microsoft platform to support .NET.

Figure 2. Cloud Foundry as Open PaaS

Cloud Foundry supports RabbitMQ for messaging, MongoDB and Redis for NoSQL, relational
databases MySQL and PostgreSQL. Cloud Foundry can be deployed on notebooks through Micro
Cloud Foundry. It is the complete version of Cloud Foundry designed to run in a virtual machine.
It can also be deployed on Private Cloud or Public Cloud. These features made Cloud Foundry as
a flexible PaaS.

Cloud Foundry components perform routing, authentication, messaging, logging, application

storage and execution, provide services and take care of application life cycle. The router routes
incoming traffic to the appropriate component, usually the Cloud Controller or a running
application on a DEA (Droplet Execution Agent) node. The User Account and Authentication
(UAA) server work with Login Server to provide identity and authentication management.
OAuth2 Server is uses as the user account and authentication server. Cloud controller and health
56 Computer Science & Information Technology (CS & IT)
manager components take care of the application lifecycle in the cloud foundry. Cloud controller
is responsible for managing the lifecycle of applications. When a developer pushes an application
to cloud foundry,
oundry, application is tar
targeting the cloud controller. Cloud controller
ontroller then stores the
raw application bits, creates a record to track the application metadata, and directs a DEA node to
stage and run the application.. Health manager m monitor
onitor applications to determine their state,
state
version,
ersion, and number of instances. Applications state may be running, stopped, or crashed.
crashed Health
manager determine
etermine applications expected state, version, and number of instances. It reconciles
the actual state of applications with their expected state. Health manager direct irects the cloud
controller
ontroller to take action to correct any discrepancies in the state of applications. The Droplet
Execution Agent manages application instances, trackstracks, started instances, and broadcasts state
messages. Application instances live inside warden arden containers. Containerization ensures that
application instances run in isolation, get their fair share of resources, and are protected from
noisy neighbours. Blob Store holds the application code, build packs, and droplets. Applications
typically depend on services like databases or third-partyparty SaaS providers. When a developer
provisions and binds a service to an application, the service broker for that service is responsible
for providing the service instance. Cloud Foundry uses a lightweight publish-subscribe subscribe and
distributed queueing messaging system for internal communication between components.
components This
internal communication performed via message bus. The metrics collector gathers metrics from
the components. Operators can use this informatio
informationn to monitor an instance of Cloud Foundry.
The application logging aggregator streams the application logs to the corresponding developers.
Cloud Foundry components are shown in Figure 3.

Figure 3. Cloud foundry components

4. CLOUDIFY
Cloudify [11] is another open PaaS cloud application manager. It automates common processes
needed to perform and to manage the applications in a cloud environment. Cloudify composed of
three main components. The components are Command line interface clien client,
t, Agents, and
Manager. Command line interface client is an executable file which is written in Python. It is
 Computer Science & Information Technology (CS & IT) 57
packaged with python and relevant dependencies in an executable file. Command line interface
client can run on Windows, Linux and Mac operating sy systems.
stems. Command line interface client
preform two tasks. First one is manager bootstrapping and another is managing applications.
Bootstrapping is the process of installing the Cloudify manager. Command line interface client
provides the user with the full set of functions for deploying and managing applications including
log and event browsing.

Cloudify Agents are responsible for managing the manager’s command execution using a set of
plugins. There is a manager side agent per application deployment and opt optional
ional agent on each
application Virtual Machine (VM). The manager side agents handle IaaS related tasks, like
creating a VM or a network, and binding a floating IP to a VM. Manager side agents can also be
used with other tools such as REST to remotely execexecute
ute tasks. The application side agents are
optionally located on application VM’s. The user can state in the blueprint which VM’s will have
an agent installed on them. The application side agents are installed by the manager side agent as
part of the VM creation
eation task. Once running, the application side agent can install plugins and
execute tasks locally. Typical tasks will be middleware installation and configuration, and
application modules deployment.

Cloudify Manager deploys and manages applications des described

cribed in blueprints. The manager’s
main responsibilities are to run automation processes described in workflow scripts and issue
execution commands to the agents. Cloudify is controlled via a REST API. The REST API
covers all the cloud orchestration and m management
anagement functions. Cloudify’s Web GUI works with
the REST API to add additional value and visibility. Cloudify uses a Workflow engine to allow
automation process through built
built-in
in and custom workflows. Workflow engine is responsible of
timing and orchestrating
ating tasks for creating or manipulating the application components. The user
can write custom workflows in Python using API’s that provide access to the topology
components.

Figure 4. Cloudify Stack

Cloudify uses different databases as data store, some of the technologies for processing and
messaging, and different servers as front end. Total stack is shown in Figure 4. Cloudify uses
elastic search as its data store for deployment state. The deploymen
deploymentt model and runtime data are
stored as JSON documents. Blueprints are stored in the elastic search and it is used as runtime
DB. Cloudify uses InfluxDB as the monitoring metrics repository. Influx provides flexible
schema for metrics and metrics metadata aass well as a query language. Cloudify stores every
metric reported by a monitoring tool into influxdb and define time based aggregations as well as
58 Computer Science & Information Technology (CS & IT)
statistic calculations. Clodify uses RabbitMQ task broker for messaging. Cloudify offers a policy
engine that runs custom policies in order to make runtime decisions about availability, service
level agreement, etc. For example, during installation, the policy engine consumes streams of
events coming from monitoring probes or tools. The policy engine analyses these streams to
decide if a specific node is up and running and provides the required functionality. Policies are
registered, activated, deactivated and deleted by the Workflow Engine. For logging purpose
logstash is used and agent play main role in processing. Nginx proxy and file server, Flask or
Gunicorn REST server, and Node.js GUI servers can be used as front end in the Cloudify.

5. OPEN SHIFT
OpenShift [12] enables us to create, deploy and manage applications within the cloud. Two basic
functional units of the Openshift are the Broker and Node servers. Communication between the
Broker and Nodes is done through a message queuing service. Broker is the single point of
contact for all application management activities. It is responsible for managing user logins,
DNS, application state, and general orchestration of the applications. Customers don’t contact the
broker directly; instead they use the Web console or CLI tools to interact with Broker over a
REST based API. Nodes are the systems that host user applications. In order to do this, the Nodes
are configured with Gears and Cartridges. A gear represents the part of the Node’s CPU, RAM
and base storage that is made available to each application. An application can never use more of
these resources allocated to the gear, with the exception of storage. OpenShift supports multiple
gear configurations, enabling users to choose from the various gear sizes at application setup
time. When an application is created, the Broker instructs a Node to create a new gear to contain
the application. Cartridges represent pluggable components that can be combined within a single
application. These include programming languages, database engines, and various management
tools. Users can choose from built-in cartridges that are served directly through OpenShift, or
from community cartridges that can be imported from a git repository. The built-in cartridges
require the associated languages and database engines to be installed on every Node.

6. OTHER PAAS
In this section we are going to give brief introduction about Stratos, Stakato and mOSAIC open
PaaS environments.

6.1. Stratos

Apache Stratos [5] is a highly-extensible PaaS framework that helps to run Apache Tomcat, PHP,
and MySQL applications, and can be extended to support many more environments on all major
cloud infrastructures. For developers, Stratos provides a cloud-based environment for developing,
testing, and running scalable applications. In Single JVM deployment model Stratos could
accommodate up to 100 cartridge instances. In a distributed deployment model Stratos could
accommodate up to 1000 cartridge instances.

6.2. Stakato

Stackato [6] is open PaaS software based on Cloud Foundry, Docker and other open-source
components. It has multi-tenancy capabilities and can be installed on internal infrastructure or
public cloud. Multi-tenancy capabilities are important because they allow us to run multiple
 Computer Science & Information Technology (CS & IT) 59
applications on the same IaaS infrastructure. Stackato allows developers to automatically package
applications into their own Docker containers and scales instances up or down on demand.
Stackato provisions all required components, including languages, frameworks and service
bindings, automates logging and monitoring, allows for automated application versioning and
rollback.

6.3. mOSAIC

mOSAIC [9] is an open-source API and platform for designing and developing multi-Cloud-
oriented applications. The architecture has been designed with open and standard inter faces. The
main goal is to provide a unified cloud programming interface which enables flexibility to build
applications across different cloud providers. The main middleware components providing
integration features are the Cloudlet, Connector, Interoperability, and Driver API. The Cloudlet
and Connector API layers facilitate the integration into the target language environment which is
used by the developers in their applications. The Driver API layer provides abstraction over
resource allocation on top of the native resource API. Interoperability API is the middleware
layer that integrates the connector API and compatible driver API implementations that could be
written in different languages. It is a remote API that follows the model of RPC with
functionalities including marshalling, request/response correlation, and error detection. Apart
from its cloud integration features, mOSAIC framework is promised to have a semantic-oriented
ontology for describing cloud resources.

7. COMPARISON OF MAJOR PAAS

This section compares the major open PaaS frameworks. Table 2 shows the basic functionality
and its corresponding AppScale, Cloud Foundry, Cloudify, and OpenShift architectural
components.
Table 2. Open PaaS Components comparison

Functionality AppScale Cloud Foundry Cloudify OpenShift

Core functionality AppController Cloud controller Manager Broker
Providing third party database Database Master Service Broker Agent Cartridge
services
Routing of incoming traffic AppLoadBalancer Router Manager REST API
Querying the state of apps AppScale Tools Cloud controller CLI client Broker
Messaging AppController Message Bus Manager Broker
Application instance AppServer Droplet Execution Agent Node
management Agent
Application state change AppLoadBalancer Health Manager Manager Broker
Containerization Database Slave Warden Agent Gear
Load balancing of user AppLoadBalancer Droplet Execution Manager Broker
requests Agent
Framework provider AppServer Blob Store Agent Cartridge

Table 3 shows the AppScale, Cloud Foundry, Cloudify, and OpenShift PaaS supported languages
(java, python, ruby), databases (MongoDB, MySQL, HBase) and frameworks (spring, rails, and
flask). In OpenShift, languages and databases are supported in the form of cartridges. User
defined cartridges are also allowed in OpenShift. Cloud Foundry provisions languages in the
60 Computer Science & Information Technology (CS & IT)
form of build packs. Users can also pick to write their own build packs. Cloudify, Cloud Foundry
and Openshift have extensible language support feature.

Table 3. Language, Database and Frameworks supported by open PaaS

Languages Databases Frameworks

AppScale Python, Java, Go, PHP Cassandra, HBase, Django, Flask,
Hypertable, MongoDB, Spring
SimpleDB, MySQL
Cloud Java, Ruby, Scala, Node.js, MonogoDB, MySQL, Spring, Rails,
Foundry Groovy, Grails, PHP, Go, PostgreSQL Grails, Play,
Python Sinatra
Cloudify Java, PHP, Ruby MySQL, MongoDB -
OpenShift Java, PHP, Ruby, Python, Perl, PostgreSQL, MySQL, Rails, Flask,
JavaScript, Node.js MongoDB Django, Drupal,
Vert.x

Table 4 shows the features support by AppScale, Cloud Foundry, Cloudify, and OpenShift
platforms.

Table 4. Open PaaS Considerable Feature Support

Features AppScale Cloud Foundry Cloudify OpenShift

Relational database support Yes Yes Yes Yes
NoSQL database support Yes Yes Yes Yes
Horizontal Scaling Yes Yes Yes Yes
Vertical Scaling No Yes No Yes
Auto Scaling Yes No Yes Yes
Spring Framework support Yes Yes No No

8. CONCLUSIONS

Cloud computing service models like Software as a Service (SaaS), Platform as a Service (PaaS),
and Infrastructure as a Service (IaaS) are introduced in this paper. PaaS is explained in detail with
the help of open PaaS packages like AppScale, Cloud Foundry, Cloudify, and OpenShift.
AppScale components are explained in table format, Cloud Foundry components are explained in
detailed with a diagram, Cloudify and OpenShift components are also explained. Stakato, Stratos
and mOSAIC open PaaS environments also explained in this paper. Comparative study is
performed among the AppScale, Cloud Foundry, Cloudify and OpenShift open PaaS componets.

ACKNOWLEDGEMENTS

This research was supported by Basic Science Research Program through the National Research
Foundation of Korea (NRF) funded by the Ministry of Education. (Grant Number: 2011-
0011507).
 Computer Science & Information Technology (CS & IT) 61

REFERENCES

[1] Daniel Nurmi, Richard Wolski, Chris Grzegorczyk, Graziano Obertelli, Sunil Soman, Lamia Youseff
& Dmitrii Zagorodnoy, (2009) “The Eucalyptus Open-Source Cloud-Computing System”, CCGrid,
pp124-131.

[2] Bunch, C., Chohan, N., Krintz, C., Chohan, J., Kupferman, J. & Lakhina, P., et al., (2010) “An
Evaluation of Distributed Datastores Using the AppScale Cloud Platform”, IEEE International
Conference on Cloud Computing.

[3] Bunch, Chris, Navraj Chohan & Chandra Krintz, (2011) “Appscale: open-source platform-as-a-
service”, UCSB Technical Report.

[4] Varma, N. M. K., Min, D. & Choi, E. (2011) “Diagnosing CPU utilization in the Xen virtual machine
environment”, In Computer Sciences and Convergence Information Technology (ICCIT), 6th
International Conference, pp. 58-63, IEEE.

[5] Pawluk, P., Simmons, B., Smit, M., Litoiu, M. & Mankovski, S. (2012) “Introducing STRATOS: A
cloud broker service”, In 2012 IEEE Fifth International Conference on Cloud Computing, pp. 891-
898, IEEE.

[6] Fortiş, T. F., Munteanu, V. I., & Negru, V., (2012), “Towards a service friendly cloud ecosystem”, In
Parallel and Distributed Computing (ISPDC), 11th International Symposium, pp. 172-179, IEEE.

[7] Hossny, E., Khattab, S., Omara, F. & Hassan, H., (2013) “A Case Study for Deploying Applications
on Heterogeneous PaaS Platforms”, In Cloud Computing and Big Data (CloudCom-Asia),
International Conference on (pp. 246-253), IEEE.

[8] Varma, N. M. K. & Choi, E., (2013) “Extending Grid Infrastructure Using Cloud Computing”, In
Ubiquitous Information Technologies and Applications, pp. 507-516, Springer Netherlands.

[9] Marpaung, J., Sain, M. & Lee, H. J., (2013) “Survey on middleware systems in cloud computing
integration”, In Advanced Communication Technology (ICACT), 15th International Conference, pp.
709-712, IEEE.

[10] D. Bernstein, (2014) “Cloud Foundry Aims to Become the OpenStack of PaaS”, IEEE Cloud
Computing, (2), 57-60.

[11] Graham, S. T. & Liu, X., (2014) “Critical evaluation on jClouds and Cloudify abstract APIs against
EC2, Azure and HP-Cloud”, In Computer Software and Applications Conference Workshops
(COMPSACW), IEEE 38th International, pp. 510-515, IEEE.

[12] A. Lomov, (2014) “OpenShift and Cloud Foundry PaaS: High-level Overview of Features and
Architectures”, Available at www.altoros.com/openshift_and_cloud_foundry_paas.html.
62 Computer Science & Information Technology (CS & IT)

AUTHORS

Name: Mohan Krishna Varma Nandimandalam

Address: B-304, DIS Lab, School of Business IT, International Building,
Kookmin University, Seoul-136702, South Korea
Education: Completed Bachelor of Computer Applications degree in 2002, Received
Master of Science in Information Systems degree in 2004 and Master of Technology in
Computer Science and Engineering in 2007 from VIT University, India. At present studying Ph.D. in
Graduate School of Business IT, Kookmin University, South Korea.

Eunmi Choi is a Professor in the School of Business IT,

Chairperson of School of Management Information Systems,
Head of Distributed Information System & Cloud Computing Lab., and
Executive Chief of Business IT Graduate School at Kookmin University, Korea,
Her current research interests include big data infra system and processing, cloud
computing, cyber physical system, information security, distributed system, SW meta-
modelling, and grid & cluster computing. Professor Choi received and MS and PhD in computer science
from Michigan State University, U.S.A., in 1991 and 1997, respectively, and BS in computer science from
Korea University in 1988.