Autonomous, Seamless and Resilience Carrier Cloud Brokerage Solution for Business Contingencies during Disaster Recovery

This paper proposes an autonomous and fault tolerant carrier cloud brokerage solution with resilient provisioning of on-the-fly cloud resources for disaster recovery management for cloud based services. The proposed solution will trigger the migration of an entire IaaS from one cloud to another without causing any disruption to the business. It also proposes an on-the-fly host selection mechanism to locate other active compute hosts without any disruptions. The solution reduces capital expenditure and provides a reliable and efficient way to access the data during disaster.

Autonomous, Seamless and Resilience Carrier Cloud Brokerage Solution for
Business Contingencies during Disaster Recovery
Sonia Shahzadi∗, George Ubakanma†, Muddesar Iqbal‡, Tasos Dagiuklas§
∗Email: [e-mail] † ‡§ Email: {ubakang, m.iqbal, tdagiuklas}@lsbu.ac.uk
∗ Swan Mesh Networks Ltd, Research and Development, London, UK
† ‡§ School of Engineering, London South Bank University, UK
Abstract—The challenge of disaster recovery management
for cloud based services is constantly evolving. The costs of
cloud service downtime in the event of disaster striking is
the subject of much international research. The key issue to
resolve is developing suitably resilient and seamless live/real-
time mechanisms for disaster recovery. In this paper, we
have implemented a proof of concept for an autonomous and
fault tolerant carrier cloud brokerage solution with resilient
provisioning of on-the-fly cloud resources. When a disaster
strikes, the proposed solution will trigger the migration of an
entire IaaS from one cloud to another without causing any
disruption to the business. In the event of non-availability
of hosts for the deployment of virtual network functions
for different business processes, an on-the-fly host selection
mechanism is proposed and implemented to locate other active
compute hosts without any disruptions. In order to evaluate the
performance of the proposed solution, we defined several use-
case scenarios for each cloud service. This proposed solution
will not only reduce the capital expenditure but also provides
a reliable and efficient way to access the data during disaster.
Keywords-Cloud Computing, Business Continuity, Infras-
tructure as a Service, Platform as a Service, Software as a
Service
I. INTRODUCTION
Disaster recovery in cloud computing has gained a lot
of attention due to its benefits that facilitate the needs
of business. Currently, many business organizations face
diverse disruptions that could affect organizational assets.
Therefore, a proactive approach is required in order to
counteract these disruptions. Mostly, these organizations
depend on Disaster Recovery (DR) services to prevent
service disruptions, because even short periods of downtime
can cause significant business losses [1]. Generally, these
DR services are expensive increasing the organizations
capital costs. The proposed solution addresses on-demand
and autonomous mechanisms to assure high availability
of services in disaster situations. When we study disaster
recovery in cloud environments different options are
available. Cloud computing depends on cloud models i.e.
private cloud, public cloud and hybrid cloud [2].
The Sendai Framework for Disaster Risk Reduction 2015-
2030 [3], was adopted at the Third UN World Conference
in Sendai, Japan, in March 2015. The Sendai Framework
outlines the need to increase investment efforts in Disaster
Reduction for Resilience. As part of this on-going research
effort we have developed a cost effective, fault tolerant and
resilient carrier cloud architecture that can be deployed at
short notice in disaster management and recovery situations.
The solution builds on the principles of live migration and
disaster recovery [2]. In order to keep the solution flexible
and accessible open source technologies and existing
standards are utilized to maximum effect in delivering a
practical, robust and extensible solution.
Cloud computing provides feasible disaster recovery
solutions due to its dynamic scalable and high availability
structure. To reduce the disaster outcomes, a multi-cloud
disaster recovery model is implemented that mange the
resources from multiple cloud providers. In this paper, we
will argue that cloud computing is an effective platform
for DR services with low costs, and it minimizes recovery
time without data loss. The reason to choose cloud based
DR solution [4]:
• Easy to deploy and manage
• Maximum flexibility
• Agility
• High availability
• Reduce capital cost
• Reduce DR solution cost
• Ease of access for business continuity
The remainder of this paper is organized as follows. Sec-
tion II describes the relevant work and state of the art.
Section III presents use cases relevant to cloud disaster
management. Section IV presents the architecture of our
proposed solution. Section V describes the implementation
of a proposed framework as proof of concept. Section
VI describes the performance evaluation while section VII
describes the conclusion of this paper and future work.
II. RELEVANT WORK
Today, as technology evolves rapidly, dynamic and re-
sponsive applications are required to support users. Develop-
ment flow of Business Continuity (BC) and disaster recovery
are given in [5] where two parts of a disaster recovery
program (i.e. building the program and choosing cloud
solution) using cloud are discussed. According to study

[4], many organizations are considering migration to cloud
computing for business continuity and a survey revealed
that only 50% of re-respondents have proper DR plans for
the whole organization, while 36% say they only have DR
plans for back-end infrastructure that will only work for data
center, not for remote offices and desktops, 7% don’t have
DR plans, but they can deploy within 6 months, 5% don’t
have DR plans, but they can deploy within 12 months and
2% also don’t have DR plans, but they can deploy within 24
months. Business Continuity and Disaster Recovery (BCDR)
plans for healthcare scenario are discussed in [6] to make
sure fast and secure access of patient data with reasonable
budget. A practical solution is implemented to deal with
disaster recovery and business continuity[7] in Portugal.
Cloud brokerage solutions can provide on-the-fly configu-
ration of existing cloud platforms at the Infrastructure as a
Service (IaaS), Platform as a Service (PaaS) and Software
as a Service (SaaS) levels [8], while supporting portability
and interoperability solutions [9]. Mostly, the cloud porta-
bility and interoperability approaches are put into practice
to overcome certain cloud platform limitations. Although,
portability and interoperability technically complement each
other, both approaches differ in many ways [9]. To reduce
the human effort, design and deployment of cloud appli-
cations is provisioned such that solutions can move from
one cloud to another, this is called portability [9]. Whereas
cloud interoperability supports different cloud services and
applications to work together and the customer is largely
unaware of this [9]. Based on the literature review, the
existing approaches can be categorized into the following
two units [9]:
A. Using existing standards
1) Cloud Data Management Interface (CDMI) specifies
how applications: create, delete, update and retrieve
data on the cloud.
2) Open Cloud Computing Interface (OCCI) supports the
deployment, monitoring and autonomic scaling. Its
API supports compute, network and storage services
[9].
3) Open Virtualization Format (OVF) for import and
export of VMs using OVF standards.
4) Topology and Orchestration Specification for Cloud
Applications (TOSCA) specify a language to a define
service and its components to deal with portability [8,
9].
B. Using open source libraries
1) Jclouds abstract the differences between multiple
cloud providers and also provide application portabil-
ity [10].
2) δ-Cloud is REST based API framework in Ruby
language that abstract the differences between multiple
cloud IaaS platforms [11].
3) Libcloud is a Python library that support extensive
cloud providers APIs [12].
4) Fog provides a high level interface to different clouds
using ruby language [13].
5) Dasein Cloud is a Java based library for compute
service access [14].
6) Simple Cloud is a PHP based library for storage, queue
and infrastructure services [15].
These open source abstraction solutions provide an interme-
diate layer for cloud management [8, 9, 16, 17, 18] while
comparisons of these solutions are given in Table I.
III. CLOUD DISASTER MANAGEMENT : USE CASES
Regardless of whether the disaster is classified as:
Natural e.g. hurricane, tornado, flood or earthquake.
Man-made e.g. infrastructure failure and cyber-attacks etc.
A Disaster Recovery (DR) plan incorporating procedures
and techniques for prevention, mitigation, managing recov-
ery is essential. Within the scope of the practical measures
for executing the DR plan, our research focuses on the
delivery of an instant and scalable approach enabling:
A. Reduced Recovery Waiting Times
For good DR services according to cost, these matrices are
used: Recovery Time Objective (RTO), Recovery Point Ob-
jective (RPO), performance and geographic separation [19].
We are developing a proof of concept utilizing our Multi-
Cloud Broker Orchestrator and Cloud Carrier Architecture
to provision instant (on the fly) deployment/restoration of
essential data center services. This enables deployment of
essential services sooner when a disaster has struck. To
enable this were centrally orchestrating all services instead
of provisioning them separately. Our goal is improving
metrics such as RTO, the maximum time to service recovery
and the RPO the maximum allowable data loss. The values
assigned to metrics will be defined on an individual basis
by problem domain and application [20]. Cloud Replication
backup approach offers short Recovery Time Objectives
(RTOs) with maximum protection for critical applications
[6].
B. Service Traffic Optimization
Our goal is to optimize traffic flows to help keep recovery
time and data losses down, while minimizing costs. Our
concept employs open source Apache libcloud to utilize a
wide range of cloud resources and to optimize in real-time,
both inter-cloud management and load balancing.
C. Backup as a Service
For good DR services according to backup, these matrices
are used: Hot Backup Site, Warm Backup Site and Cold
Backup Site [19]. Multi-Cloud Broker Orchestration and
Cloud Carrier Architectures can play an increasing role in
delivering Backup as a Service. Ensuring that clients can not

Table I
COMPARISON OF CLOUD ABSTRACTION SOLUTIONS
Features Jclouds δ-Cloud Libcloud Fog Dasein Cloud Simple Cloud
Type Library Framework Library Library Library Library
Database N N/A N N N/A N/A
Supported IaaS Y Y Y Y Y Y
Supported PaaS Y Y Y N/A N/A N/A
Multi-IaaS Support Y Y Y N/A N/A Y
Multi-Cloud Support Y Y Y N/A Y Y
DNS N N/A Y Y N/A N/A
License Apache License 2.0 Apache License 2.0 Apache License 2.0 MIT License Apache License 2.0 Open BSD License
Documentation Good Good Good N/A Very less to no documentation. Good
Supported CSPs 30 17 60 43 N/A N/A
Compute Y Y Y Y Y N/A
Network N Y Y N/A Y N/A
Storage Y Y Y Y Y Y
Amazon EC2 Support Y Y Y N Y Y
Programming Language Java Ruby Python Ruby Java PHP
Platform Integration Maven Drivers Drivers N/A N/A N/A
EBS Storage Support Y No, but available in road-map. N N/A N/A Y
Container N N/A Y N N/A N/A
CDN N N/A Y Y N/A N/A
Load Balancing Y Y Y N N/A N/A
Y=Yes; N=No; N/A=Not Available;
only recover their cloud infrastructure and services on the fly
when disaster strikes; but also be assured of reliable access
to backup data resources as vital services are re-stored [21].
Based on the use-cases discussed above, to provision cloud
contingency services during emergency and dis-aster situa-
tions, we consider following solutions to deliver the cloud
infrastructure as a service.
During a natural disaster when communication infra-
structure is destroyed, it can also effect regional cloud data-
centers located within that specific zone. This can partially
or fully disrupt business services, resulting in potential losses
of millions. The Cloud is a promising solution to providing
on-demand provisioning services between different regions.
As a cloud manages all these resources through a central
location. Therefore, as shown in Figure 1, if regional a
cloud goes down or collapses for example in Zone B,
then all the user requests will be redirected to Zone A
and vice-versa. This may effect Zone A performance due
to extra load. During disaster recovery situations every
second counts and restoring cloud resources can become a
matter of survival for a particular business. This requires an
autonomous seamless and resilience carrier cloud broker-age
solution as shown in Figure 2, where multiple clouds are
connected to a single cloud brokerage solution which can
provision resources to accommodate different IaaS requests.
Cloud service providers can provide scalable resources to
accommodate the requirements within minutes. The entire
process is required to be initiated dynamically. In order to
achieve this, we have proposed a real-time cloud brokerage
that provides seamless, autonomous (self-service and self-
managed) resource provisioning for a contingency cloud
to replicate the destroyed IaaS during a natural disaster.
The proposed cloud brokerage will automatically trigger the
deployment of contingency cloud resources and redirect the
user request to the alternative cloud services.
Figure 1. Cloud Service Scenario for Disaster Management
IV. PROPOSED ARCHITECTURE
There are multiple solutions of data recovery in cloud
computing that depends on user requirements and IT budget
or you can mix multiple approaches according to your
unique scenario. As, we have merged two approaches, one
is for rapid data recovery while other is for cloud controller
lost. We have proposed a carrier cloud brokerage solution
for federated cloud portability and to provide resilient in-
frastructure services on demand. It provides synchronization
among multiple clouds platforms through a central broker.
The proposed carrier cloud brokerage solution will on-the-
fly identify and select the best available resources in the
region and migrate the load to it. Multiple options for cloud
services available in a particular zone are given to the users
to select resources based on their preferences. If one server
in a specific zone collapses, the user will automatically be
connected to an-other server transparently.

Figure 2. Muti-Cloud Brokerage Solution
We have developed a brokerage solution using libcloud SDK
for on-the-fly resource provisioning of cloud services during
disaster recovery. The libcloud SDK uses native cloud APIs
thus is compatible with a variety of cloud platforms. Apache
libcloud is an open-source project that is helpful for inter
cloud resource management [12]. Table II explains different
terminologies used in libcloud SDK [12]. The architecture
of the proposed framework is given in Figure 3.
Figure 3. Carrier Cloud Architecture
V. IMPLEMENTATION
We have designed and implemented an autonomous and
resilient cloud brokerage solution to provide on the fly
IaaS resource provision during disaster management and
recovery. We have implemented test-beds using open source
cloud software. We have developed and implemented a cloud
brokerage solution that can connect to multiple clouds. To
implement the cloud testbed, controller, compute, network
and storage nodes are used. Controller node is used to
control all cloud services and activities. Compute nodes
provide computing services as all VMs will run on compute
servers. Network nodes will provide networking services to
VMs. While storage nodes are used to store and retrieve
data. We have used multiple compute and storage nodes to
replicate seamless services in case of any failure.
We have assumed that each compute node is hosting virtual
resources in a separate zone. This is in case one compute
node in a particular zone fails or collapses as a result of
natural disaster. The cloud brokerage will auto-initiate the
recovery process through the cloud brokerage to identify and
locate the suitable resource to migrate the business services.
The storage services will also be replicated using the backup
storage data to store user data on other nodes. In case of any
storage server failure, users can still retrieve their data from
other nodes.
VI. PERFORMANCE EVALUATION
In order to evaluate the performance of the proposed
framework, a benchmarks approach is used to measure
the performance. The benchmarks framework is the col-
lective experiences of the cloud services to observe cloud
capacities and response while accommodating unforeseen
circumstances. As, every cloud has its own resource lim-
itations and we cannot allocate more resources than its
capacity. Therefore, benchmarking is the best practice to
measure cloud performance for cloud users and provides
an awareness for underloaded or overloaded exposure. It
facilitates the process of planning, monitoring and evaluation
of cloud services in an appropriate manner across the cloud
users. The main objective of this benchmarking is to validate
the cloud services, identify the gaps between cloud services
and replicate the cloud services during disaster recovery.
Since we have assumed that each compute host is located in
a separate zone, if any cloud compute host server is down
in case of disaster, the requests will be automatically redi-
rected to another available compute host server via dynamic
VM consolidation. This scheduler service will automatically
catch up the other active compute host on-the-fly. Similarly,
cloud storage servers can also be replicated and replaced
using an on-the-fly mechanism during the disaster recovery
process. As, cloud computing platforms support different
disaster recovery approaches. We have implemented the
following two type of approaches that will support a DR
plan in a cloud environment.
A. VM Migration for Compute Service
We have implemented nova service with multiple compute
servers to support VM migration in the disaster environment.
This nova service will provide compute service to the cloud
to manage the instance life cycle and scheduling [22].

Table II
TERMINOLOGY IN LIBCLOUD
Terminology Description
Node Represents a virtual machine/ virtual
server.
Node Image Represents an operating system.
Node Size Represents the hardware configuration
of the virtual server, including CPU,
RAM and Disk.
Node Location Represents a physical location of vir-
tual servers. In our case, VM availabil-
ity zone is nova.
Node State Represents a node state; either node is
running, terminated or rebooting etc.
Multiple compute servers are used to facilitate the migration
services from one region to another in case of disaster. When
one compute server goes down, VM will migrate to another
compute host server. This DR mechanism is fully transparent
and scalable during a disaster.
We have performed the following tests to measure the perfor-
mance of server migration in case of disaster management.
1) Performance measurement of boot and migrate server:
In this experiment, we have deployed multiple compute
servers to deal with the disaster situation and these are
compute servers are available in nova compute zones. We
have launched a server on an available compute node. When
we detect any disaster then we will migrate this server to
another available compute server in the availability zone.
Once the VM is migrated to a new compute server, then
this experiment will confirm the resize of the VM on new
compute host and later delete the VM.
We have measured the performance of this scenario with
five iterations with four atomic actions i.e. nova.boot server,
nova.migrate, nova.resize confirm and nova.delete server.
As shown in Figure 4, this performance experiment results in
a maximum time of 65.573 sec to complete this task, while
49.502 sec is the minimum time to complete this experiment.
Figure 4. Boot and Migrate Server
B. Replicas for Storage Service
We have implemented swift services with a cloud to store
unstructured data. This storage service is fully distributed
and provides high availability and scalability to store and
get data on demand.
Swift consists of proxy server and storage server [22]. Proxy
server will forward the data to the storage server and the
storage server will store that data. Ring will decide the
optimal storage server to store data. For example, a user
requests data storage on cloud swift. All swift replica storage
servers will store the data. When users request that the data,
an appropriate storage server will respond to the user with
that data. A replicator is used to recover data in case of
disaster recovery. This replicator detects the lost data from
storage server and replicates this data to another storage
server as a temporary measure. When the damaged server
is replaced with a new server then the replicator replicates
the data on a fresh server and removes the data from the
temporary server. The following servers are also part of swift
service.
- Object Server: is responsible for storage, retrieval and
deletion of objects [22].
- Container Server: is responsible for listing of objects [22].
- Account Server: is responsible for listing of containers
[22].
we have performed the following tests to measure the
performance of this storage service in case of disaster
management.
1) Performance measurement of storage container cre-
ation with objects and download objects: In this experiment,
we have used multiple storage nodes to deal with disaster
emergency problems. On these storage nodes, we have
created a storage container to store unstructured data on
multiple objects. These objects are responsible for storage,
retrieval and deletion of data. After creation of these objects
we have also measured the download time of the data.
We have measure the performance of the proposed so-
lution with six iterations with these three atomic ac-
tions i.e. swift.create container, swift.create 5 objects and

swift.download 5 objects. As, Figure 5 represents this per-
formance experiment results with maximum time is 2.448
sec while 0.533 sec is minimum time to complete this
experiment.
Figure 5. Container creation with objects and download objects
2) Performance measurement of storage container cre-
ation with objects and then delete all : In this exper-
iment, we have measured the performance of the pro-
posed solution as per container creation with specific num-
ber of objects and then delete both things (i.e. object
and container) but make sure before deleting the con-
tainer, you need to delete objects where data are stored.
We have performed four iterations with the same atomic
actions (i.e. swift.create conatiner, swift.create 5 objects,
swift.delete 5 objects and swift.delete conatiner). As, Fig-
ure 6 represents this performance experiment resulting in a
maximum time is 1.329 sec while 0.787 sec is minimum
time to complete this experiment.
Figure 6. Container creation and deletion with objects
VII. CONCLUSION AND FUTURE WORK
Business continuity is a major requirement for all orga-
nizations in case of disaster data losses during disasters can
create a huge loss for the business including financial and
reputation damage. Various organizations often invest certain
amounts to validate data recovery quickly during disaster.
However, today cloud computing offers flexible solutions to
fulfill the business needs according to business requirement.
In this paper, an autonomous seamless cloud solution is
presented for data recovery during disaster management.
Currently, this solution is implemented for a private cloud
that fit for the single organization requirements. We have
per-formed various tests to validate our solution which sig-
nificantly provides resilient and robust cloud performance.
In future work, we can move to PaaS with light-weight
technology like container (docker) which is faster than
traditional VM’s.
NOTES
ABBREVIATIONS
Business Continuity and Disaster Recovery (BCDR),
Cloud Data Management Interface (CDMI), Disaster Re-
covery (DR), Recovery Time Objective (RTO), Recovery
Point Objective (RPO), CSP (Cloud Service Provider), In-
frastructure as a Service (IaaS), Open Cloud Computing
Interface (OCCI), Open Virtualization Format (OVF), Topol-
ogy and Orchestration Specification for Cloud Applications
(TOSCA), Platform as a Service (PaaS), Software as a
Service (SaaS), Elastic Block Store (EBS), Domain Name
System (DNS), Content Delivery Network (CDN), Business
Continuity (BC).
CONFLICT OF INTEREST
The authors declare that they have no conflict of interest.
AUTHORS ’ INFORMATION
Sonia Shahzadi received her BS and MS degrees from
University of Gujrat, Pakistan, in 2013 and 2016 respec-
tively. She is currently associated with Swan Mesh Networks
Ltd, Research and Development, London, UK. Her research
interests include Cloud Computing and Mobile Edge Com-
puting.
Muddesar Iqbal is Senior Lecturer in Mobile Computing in
the Division of Computer Science and Informatics, School
of Engineering. He won an EPSRC Doctoral Training Award
in 2007 and completed his PhD from Kingston University
in 2010 with a dissertation titled Design, development,
and implementation of a high-performance wireless mesh
network for application in emergency and disaster recovery.
He has been a principal investigator, co-investigator, project
manager, coordinator and focal person of more than 10
internationally teamed research and development, capacity
building and training projects. He is an established re-
searcher and expert in the fields of: mobile cloud computing
and open-based networking for applications in Education,
disaster management and healthcare; community networks;

and smart cities. His research interests include 5G network-
ing technologies, multimedia cloud computing, mobile edge
computing, fog computing, Internet of Things, software-
defined networking, network function virtualization, quality
of experience, and cloud infrastructures and services.
George Ubakanma joined London South Bank University
(LSBU: then South Bank Polytechnic) in 1992 as a Lecturer
in Operating Systems and Networking. He is currently
a Senior Lecturer, specializing in Database Management;
Business Intelligence Architecture; Systems Analysis and
Design. Current departmental duties also include the role
of Course Director for several of the Postgraduate (MSc)
courses in the Department of Informatics. As well as the
course management and student facing responsibilities that
the Course Director role brings, He also maintains an active
cross-Faculty role as a Departmental and Faculty Academic
Integrity Officer. He is currently Co-Chair for LSBU’s
Academic Integrity Co-ordinators(AIC), this role requires
regular cross-Faculty/University contact, together with the
research and dissemination of best practice with AIC’s from
all faculties as well as the University Registrar’s Team.
He has developed and co-authored various MSc courses
for the Department of Informatics (previously: School of
Computing & Mathematics). He has been course director
for several MSc courses, as well as managing the BCS
Examinations at LSBU. His course director’s role has also
required undertaking a commitment to conduct recruitment
and marketing responsibilities overseas, particularly in India
and South East Asia, as well as in Africa, particularly in
Nigeria and Ghana. He is currently actively involved in the
restructuring of the Informatics department’s postgraduate
and undergraduate course provision. He has also undertaken
the role of internal examiner/reviewer for the validation of
other LSBU courses. He is also an external examiner at other
UK universities and colleges.
Tasos Dagiuklas is a leading researcher and expert in the
fields of Internet and multimedia technologies for smart
cities, ambient assisted living, healthcare and smart agri-
culture. He is the leader of the SuITE research group at
the London South Bank University where he also acts as
the Head of Division in Computer Science. Tasos Dagiuklas
received the Engineering Degree from the University of
Patras-Greece in 1989, the M.Sc. from the University of
Manchester-UK in 1991 and the Ph.D. from the University
of Essex-UK in 1995, all in Electrical Engineering. He has
been a principle investigator, co-investigator, project and
technical manager, coordinator and focal person of more
than 20 internationally R&D and Capacity training projects
with total funding of approximately 5.0m from different
international organizations. His research interests include
Smart Internet Technologies, Media Optimization across
heterogeneous networks, QoE, Virtual Reality, Augmented
Reality and cloud infrastructures and services.
REFERENCES
[1] T. Wood, E. Cecchet, K. K. Ramakrishnan, P. J.
Shenoy, J. E. van der Merwe, and A. Venkataramani,
“Disaster recovery as a cloud service: Economic ben-
efits & deployment challenges.,” HotCloud, vol. 10,
pp. 8–15, 2010.
[2] P. Kokkinos, D. Kalogeras, A. Levin, and E. Varvari-
gos, “Survey: Live migration and disaster recovery
over long-distance networks,” ACM Computing Surveys
(CSUR), vol. 49, no. 2, p. 26, 2016.
[3] “Sendai Framework for Disaster Risk Reduction
20152030.” http://www.unisdr.org/files/43291
sendaiframeworkfordrren.pdf. Accessed: 2017-11-02.
[4] “Cloud-Based Disaster Recovery Emerging as Top
IT Priority, White Paper.” https://www.vmware.
com/content/dam/digitalmarketing/vmware/en/pdf/
whitepaper/cloud/idg-dr-whitepaper-0515-final.pdf.
Accessed: 2017-11-02.
[5] “Cloud Disaster Recovery: Public, Private or Hybrid
Cloud Solutions Supporting Disaster Recovery, White
Paper.” https://www.datalink.com/getattachment/
a00eea92-bde3-4033-92e1-7e31dfd8f504/
Cloud-Disaster-Recovery.aspx. Accessed: 2017-
11-02.
[6] “Best Practices in Healthcare IT Disaster
Recovery Planning, White Paper.” https:
//www.cleardata.com/wp-content/uploads/2014/08/
Best-Practices-Healthcare-IT-Disaster-Recovery-Planning.
pdf. Accessed: 2017-11-02.
[7] A. Prazeres and E. Lopes, “Disaster recovery–a project
planning case study in portugal,” Procedia Technology,
vol. 9, pp. 795–805, 2013.
[8] F. Fowley, C. Pahl, and L. Zhang, “A comparison
framework and review of service brokerage solutions
for cloud architectures,” in International Conference on
Service-Oriented Computing, pp. 137–149, Springer,
2013.
[9] D. Petcu and A. V. Vasilakos, “Portability in clouds:
approaches and research opportunities,” Scalable Com-
puting: Practice and Experience, vol. 15, no. 3,
pp. 251–270, 2014.
[10] “Jclouds.” http://jclouds.apache.org/. Accessed: 2017-
11-02.
[11] “δ-Cloud.” http://deltacloud.apache.org/. Accessed:
2017-11-02.
[12] “Libcloud.” http://libcloud.apache.org/. Accessed:
2017-11-02.
[13] “Fog.” http://fog.io/. Accessed: 2017-11-02.
[14] “Dasein.” http://dasein-cloud.sourceforge.net/. Ac-
cessed: 2017-11-02.
[15] “Simple Cloud.” http://addhrere:/. Accessed: 2017-11-
02.
[16] F. Meireles and B. Malheiro, “Integrated management

of iaas resources,” in European Conference on Parallel
Processing, pp. 73–84, Springer, 2014.
[17] “Cloud abstraction API’s.” https://sites.google.
com/site/jmathaiy/references-and-learning-guide/
cloud-abstraction-apis. Accessed: 2017-11-02.
[18] “Multi-cloud.” https://medium.com/@anthonypjshaw/
multi-cloud-what-are-the-options-part-1-low-level-abstraction-libraries-ce500f29120f.
Accessed: 2017-11-02.
[19] K. B. Nayar and V. Kumar, “Benefits of cloud com-
puting in education during disaster,” in Proceedings
of the International Conference on Transformations in
Engineering Education, pp. 191–201, Springer, 2015.
[20] O. H. Alhazmi, “A cloud-based adaptive disaster recov-
ery optimization model,” Computer and Information
Science, vol. 9, no. 2, p. 58, 2016.
[21] R. Jaluka, D. Meliksetian, and M. Gupta, “Enterprise
it as a service: Transforming the delivery model of
it services,” in Cloud Computing in Emerging Mar-
kets (CCEM), 2016 IEEE International Conference on,
pp. 32–39, IEEE, 2016.
[22] “OpenStack.” www.openstack.org/. Accessed: 2017-
11-02.