Cloudware Technologies - Guide to Cloud Computing for Business and Technology Managers - page 310

15.2.2 Platform Virtualization
Platform virtualization allows to run any desired operating systems and
applications in virtual environments. There are two different models: full
virtualization and paravirtualization. Both solutions are implemented on
the basis of a virtual machine monitor or hypervisor. The hypervisor is a
minimalistic meta-operating system used for distributing the hardware
resources among the guest systems and for access coordination. A type-1
hypervisor is built directly on top of the hardware; a type-2 hypervisor runs
under a traditional basic operating system.
Full virtualization is based on the simulation of an entire virtual computer
with virtual resources, such as CPU, RAM, drives, and network adapters,
including its own BIOS. Since the access to the most important resources, such
as the processor and the RAM, is passed through, the processing speed of the
guest operating systems nearly equals the speed to be expected if there was no
virtualization. Other components, for example, drives or network adapters, are
emulated. While this decreases the performance, it allows to run unmodified
guest operating systems. Paravirtualization does not provide an emulated
hardware layer to the guest operating systems, but only an application inter-
face. For this purpose, the guest operating systems need to be modified
because any direct access to hardware must be replaced by the corresponding
hypervisor interface call. This is also referred to as hyper calls (just like system
calls), which are used by the applications to call functions in the operating sys-
tem kernel. Since this approach allows the guest system to participate actively
in the virtualization (at least to some extent), a higher throughput than with
full virtualization can be obtained, especially for I/O-intensive applications.
Examples of full virtualization are the VMware products or, specifically for
Linux, the Kernel-based Virtual Machine (KVM). Under Linux, mostly Xen-
based solutions are used for paravirtualization. They play an important role,
particularly in the realization of the Amazon Web Services.
15.2.3 Storage Virtualization
Cloud systems should also offer dynamically scalable storage space as a
service. In this context, storage virtualization boasts a number of advan-
tages. The fundamental idea of storage virtualization is to separate the data
store from the classical file servers and to pool the physical storage systems.
Applications use these pools to dynamically meet their storage require-
ments. For the data transfers, a special storage area network (SAN) or a local
company network (LAN) is used. Data for cloud offerings are mostly avail-
able in the form of Web objects that can be retrieved or manipulated over the
Internet. An additional abstract administration layer is interposed between
the clients and the storage landscape so that the representation of a datum
is decoupled from its physical storage. This has a variety of advantages with
respect to data management and access scalability.