1.2. El recurso efectivo en la jurisprudencia de la Corte Interamericana de Derechos
1.2.1. Deber de los Estados de suministrar recursos efectivos
following sections.
2.11
Classification of Cloud deployment models
In this dissertation, the categorization of Cloud services and deployment models have been used to reflect our current understanding of the technology. Cloud has come a long way within a few years, and the classification is done based on the recent devel- opments scenarios.
Cloud is a centralized system; all physical resources are kept in a data center. Sev- eral models have been proposed for Cloud services and resource deployment. There are several taxonomies proposed for Cloud services, platforms, Infrastructure-level, Inter- operability and so on. Unfortunately, those works are almost a decade old [134, 142– 145]. The Cloud has evolved over the years, and our understanding of the technology has developed. While some of the concepts remain the same, some other concepts have changed.
In 2011, NIST categorized the Cloud deployment models as private, community, public, and hybrid [2]. Those were based on the earlier understanding of the Cloud deployment strategies. It was prevalent among the Cloud deployment models for sev- eral years. However, with time the understanding of deployment models have been changed.
In this dissertation, Cloud is divided into three categories based on deployment models; they are public, private and in-house Cloud. Table 2.2 provides a compar- ison of the three deployment models. This classification is made by analyzing the current Cloud deployment strategies of organizations. Those deployment models are discussed in the following sections. Originally this dissertation was submitted with this classification in September 2017.
In January 2018, VMware released a new classified of the Cloud deployment model [146]. This classification contains three models; hosting, private, and public. Community and hybrid, two deployment models, considered earlier in the 2011 NIST classification have vanished from the 2018 VMware classification.
The classification presented in this dissertation is similar to that offered by VMware. All three deployment model concepts are very similar to the three models proposed by VMware. The only difference is that in this dissertation one of the categories is named
2.11. CLASSIFICATION OF CLOUD DEPLOYMENT MODELS 38
the ‘in-house’, whereas in VMware technical report the same category is being called the ‘hosting’. Other than that, both categories embody the same concept.
Thus, the deployment models released by a leading virtualization enterprize few months after the original submission of this dissertation agrees with the new deploy- ment models presented in this dissertation. The three deployment models are discussed next.
Table 2.2: Comparison of characteristics of Cloud deployment models [3, 4].
Characteristics Types
Public Private In-house
Dedicated Multiple
customers Single customer Infrastructure
managed
Third party provider
Customer or Third
party provider Customer Infrastructure located Off-premise On-premise or
Off-premise On-premise Software update cycle Third party
provider
Customer or Third
party provider Customer Modification of
services
Not allowed by
customer Allowed by customer Service configuration Highly
standardized Mostly customizable Fully customizable Application security
requirement Usually low Medium to high High Application resource
requirement Usually High High to moderate Moderate to low
Access policy Un-trusted Trusted
Fault tolerance and recovery scheme
Third party
provider Custom scheme possible Application
performance tuning
Third party provider
Customer or Third
2.11. CLASSIFICATION OF CLOUD DEPLOYMENT MODELS 39
2.11.1
Public cloud
Public Cloud is the most common form of Cloud available today; it offers VM services to all users [29]. A user can rent computing, storage, and data transfer services from the Cloud provider [139]. The VMs are rented out in the pay-as-you-go model; a user can rent a unit of resource for any duration.
In a public Cloud, VMs from multiple tenants usually share the same physical server. Public Cloud users usually have fewer security requirements, and the resource units are relatively cheaper. Users can easily rent VMs through web portals and use them.
Most deployed applications are interactive and have low latency and throughput requirements, like the Web and game servers. Deployment of scientific application is less common in public Cloud. However, nowadays public Clouds has started offering specialized services for high resource intensive applications. Some examples of public clouds include the Amazon Elastic Cloud (EC2), Blue Cloud by IBM, Sun Cloud,
Google AppEngineand Windows Azure [147].
2.11.2
Private cloud
Private Cloud applies strict security rules and regulations [2, 29]. Extra care is taken to ensure the quality of services like, latency, and throughput. Primary users are the enterprises and departments that require greater computing power to run scientific ap- plications. Such an application requires greater computing power compared to the applications on public Cloud. However, they have a shorter lifespan compared to the applications of the public Cloud.
For example, a company may need to simulate massive chemical reactions for a new product [148, 149]. Similarly, a biology department may require working with a significant amount of biological data [150]. Such applications need high-performance computing power. The resource requirement for parallel tasks is different from that of the ordinary web servers. Furthermore, there can be sensitive data that requires more security compared to that offered by the public Cloud. The scientific data are confidential, and it needs to be secured from unauthorized access.
The private Cloud offers a higher level of security and SLA for sensitive data and applications. For example, they make sure that VMs of only one user resides on a server and VMs of no other users are on that server. Private Clouds are usually smaller