Information Technology Reference
according to Penguin Computing, is a high-performance, scalable, on-demand HPC as a
Service solution with high-density compute nodes and high-speed storage that offers users a
persistent, load on demand, virtualized compute environment (Penguin Computing, 2010).
It is likely that more cloud providers will be moving into this computing area in the future
as more people from the scientific community begin to look at this computing service for
solutions to their problems.
V. E CONOMICS , F LEXIBILITY AND G REEN C REDENTIALS
Economics, simplification and convenience of the way computing-related
services are delivered seem to be among the main drivers of cloud computing (Erdogmus,
2009). Many people see huge potential of the technology in reducing the cost of IT to
organizations and freeing them from the expense and hassle of having to install and maintain
applications locally (Leavitt, 2009). Providing IT services in the cloud shifts much of this
expense to a pay-as-you-go model and consequently offers significant cost advantages
according to one view (Lin, Fu, Zhu and Dasmalchi, 2009). Furthermore, a great proportion
of the costs of running an IT infrastructure comes from electricity consumption which is
needed to run hardware (e.g., PCs, servers, switches, backup drives) and cooling which is
needed to reduce the heating generated by the hardware. Cloud computing is likely to reduce
expenditure in this area and also reduce labor-related costs, as less people (e.g., technicians)
than before will be required to run a cloud-based IT infrastructure.
The cost advantage of cloud computing is not just related to how much cloud users can
save by not buying and installing hardware and software and using less power. Users of cloud
computing are more likely to significantly reduce their carbon footprint. Research suggests
that ICT is already responsible for 2% of global carbon emissions, and that its relative share
will increase further. In the UK, for example, increasingly stringent regulations (such as the
Carbon Reduction commitment and EU Energy Using Products Directive) are likely to put
pressure on educational establishments to make ICT more sustainable (James and
Hopkinsons, 2009). In an environment where there is increasing concern about institutions'
carbon footprint and energy costs, virtualized services (such as those offered by cloud
computing) may become especially appealing (Katz, 2008).
VI. T HE G ROWING A PPEAL OF C LOUD C OMPUTING FOR HPC
Until recently, high performance computing has not been a good candidate for cloud
computing due to a number of factors such as its requirement for tight integration between
server nodes via low-latency interconnects and high-speed networking (Shainer et al 2010).
For example, the performance overhead associated with host virtualization, a prerequisite
technology for migrating local applications to the cloud, quickly erodes application scalability
and efficiency in an HPC context which often involves sending messages back and forth
many times per second, a process that is likely to increase the possibility of latency (Niccolai,
2009). However, new virtualization solutions that use KVM (kernel-based virtual machine)