Energy-Aware Policies in Ubiquitous Computing Facilities - Cloud Computing with e-Science Applications - page 269

computing is emerging as the dominant computer platform for scalable
online services.
Thus, the wireless body sensor networks (WBSNs) will be connected not
only at the node level but also through a personal digital assistant (PDA)
or smartphone to the cloud. Part of the data processing and storage will be
local to the node, while another part will be communicated and processed
in the cloud, depending on the application, on the state of the batteries,
and on security or privacy requirements of the information. This comput-
ing environment where the mobile client utilizes mobile network services
to communicate with the cloud through the Internet is usually known as
mobile cloud computing (MCC) [1].
Recent research has focused on developing energy efficiency policies at
the data center level. Some policies have been detected but not successfully
proposed as they lack consideration of the global power consumption of
the system. They do not take into account that the agents involved in the
problem are heterogeneous. Therefore, the energy cost of performing part of
the processing in any of the different abstraction layers, from the node to the
data center, should be evaluated.
Our proposal develops global energy optimization policies that start from
the design of the architecture of the system, with a deeper focus on data
center infrastructures, and take into account the energy relationship between
the different abstraction layers, leveraging the benefits of heterogeneity and
application awareness.
12.2 Related Work
For decades, data centers have only focused on performance, defined as
speed. Examples include the TOP500 list of the world's fastest supercomputers
(http://www.top500.org), which calculates speed as floating-point operations
per second (FLOPS), and the annual Gordon Bell Awards for Performance
and Price/Performance at the Supercomputing Conference (http://www.
supercomp.org). However, raw speed has increased tremendously over the
past decade without relative and proportional energy efficiency. In 2007,
although there had been a 10,000-fold increase in speed since 1992, perfor-
mance per watt was only improved 300-fold and performance per square
foot only 65-fold [2].
This huge performance improvement is mainly due to increases in three
different dimensions: the number of transistors per processor, the operating
frequency of each processor, and the number of processors per system.
Collectively, these factors yield an exponential increase in power consump-
tion of data centers that is not sustainable. The focus on just speed has
let other evaluation metrics go unchecked. Data centers consume a huge