Evaluation Criteria to Run Scientific Applications in the Cloud - Cloud Computing with e-Science Applications - page 5

1.3.1 Private Cloud
A private cloud is basically the same as owning and maintaining a tradi-
tional cluster, where the user has total control over the infrastructure and
can configure the machines according to need. One big issue in a private
scenario is the absence of instant scalability, as the capacity of execution
is limited to the physical hardware available. Moreover, the user needs to
have access to facilities to maintain the machines and is responsible for the
energy consumption of the system. Another disadvantage is the hardware
maintenance; for example, if a machine has physical problems, the user is
responsible for fixing or replacing it. A case for which the private cloud is
recommended is if the application uses confidential or restricted data; in this
scenario, the access control to the data is guaranteed by the user's policies.
The weakness of this model is the absence of elasticity and the need for up-
front costs. Building a private cloud for scientific applications can be consid-
ered the same as buying a cluster system.
1.3.2 Community Cloud
In a community cloud, the users are members of one organization, and this
organization has a set of resources that are connected to resources in other
organizations. A user from one of the organizations can use the resources
of all other organizations. The advantage of this model is the provision
of access to a large set of resources without charging because the remote
resources belong to other organizations that form the community and not
to a provider. In other words, the pay-per-use model may not be applicable
to this type of cloud. One disadvantage of the model is the limited number
of resources; they are limited to the number of machines that are part of the
community cloud. The interconnection between all the members constitutes
a bottleneck for the application's execution. If the application needs more
machines than are available in single site (a single member), the machines
need to be allocated among two or more members.
All the community members need to use the same cloud platform; this
demands an effort to configure all the machines, and it is necessary to
have personnel to maintain the machines. The community model is recom-
mended for research groups that are geographically distributed and want to
share the resources among them.
1.3.3 Public Cloud
In a public cloud, the infrastructure is provided by a company, the provider.
The advantage in this case is the access to an unlimited number of computa-
tional resources, where the user can allocate and deallocate them according
to demand. The pay-per-use billing model is also an advantage because the
user has to spend money only while using the resources. Access to up-to-date