Principles, Methodology and Tools for Engineering Cloud Computing Systems - Advanced Design Approaches to Emerging Software Systems - page 259

environment is the possibility of storing different
versions of the components code, for example to
roll back to a previous version if some error is
detected after the last component update.

and storage, networking configuration, firewall,
security, etc. Other commercial services such as
GoGrid 4 , Flexiscale 5 follow similar approaches to
deploy cloud applications on their infrastructure,
offering proprietary ways for packaging and con-
figuring the cloud applications.
Similarly, Windows Azure 6 , as a PaaS Cloud
framework and provider, relies for example on
service models. These enable services to be de-
scribed as distributed entities: clients can specify
the interfaces exposed by services, communication
end points and channels, roles, whether web for
front-end communication or worker for back-
ground processing, and various other configura-
tion data which must be provided when hosting a
service. In addition, similarly to IaaS platforms,
because each role will run its own virtual machine
instance, different hardware requirements may be
allocated at different costs.
However, as cloud based offerings become
increasingly popular, the complexity and scale
of services deployed in the cloud means that ad-
ditional degrees of flexibility and control over
the provisioning process must be provided to the
customer. Services may incorporate an increas-
ingly wide range of components, data or applica-
tion resources and dependencies and constraints
may exist between these resources that must be
clarified when deploying, migrating and resizing
services in a cloud.
Service provider will also wish to describe
how the service responds to load variations and
faults. In order to minimize over provisioning and
optimize the use of resources, there must exist
means of specifying the adjustment of service
capacity or configuration throughout the entire
service lifetime according to application state and
workload and communicating this information to
the infrastructure provider.
These concerns highlight the fact that service
definition cannot be solely focused on the initial
deployment of fixed size application instances.
There must exist an ability to for a client to describe
the overall service architecture of applications

2.2 Service Definition

Having their service defined, providers are ready
to deploy their application service, i.e., a clear
specification of the requirements of this service
must be communicated to the underlying platform.
This includes overall requirements such as the
minimum hardware profile required by the differ-
ent service components, in terms of CPU, memory
and so on, alongside required application data such
as a virtual disk, ISO image, or application code
and libraries. This Section is unavoidably related
with Section 2.4, since service definition includes
mechanisms for controlling service scalability.
Most Cloud platforms, IaaS, PaaS, or SaaS,
will provide means for a client to define service
requirements and configuration data typically
in the form of a deployment descriptor. This
may be via a graphical user interface such as
Amazon's Web Service Management console 2 ,
or as a separate document with a clear specified
syntax, semantic and structure. Upon deployment
the provisioning framework binds the necessary
infrastructure capabilities declared in the graphi-
cal user interface or descriptor to the application,
in the form of hardware resources, such as CPU
cores or memory (which is the usual case in IaaS
clouds), or alternatively software dependencies
(usually for PaaS).
IaaS users will describe upon submission their
hardware and software requirements via a suitable
interface or descriptor. Amazon's EC2 3 offers
a Xen (Barham et al. 2003) based proprietary
format called an Amazon Machine Image (AMI)
for cloud applications. Pre-configured or custom
built images can then be deployed on Amazon's
cloud via their API or management interface, at
which time the user defines the preferred hardware
configuration in terms of CPU cores, memory