ArticlePDF Available

Abstract and Figures

The implementation of cloud computing has attracted computing as a utility and enables penetrative applications from scientific, consumer and business domains. However, this implementation faces tremendous energy consumption, carbon dioxide emission and associated costs concerns. With energy consumption becoming key issue for the operation and maintenance of cloud datacenters, cloud computing providers are becoming profoundly concerned. In this paper, we present formulations and solutions for Green Cloud Environments (GCE) to minimize its environmental impact and energy consumption under new models by considering static and dynamic portions of cloud components. Our proposed methodology captures cloud computing data centers and presents a generic model for them. To implement this objective, an in-depth knowledge of energy consumption patterns in cloud environment is necessary. We investigate energy consumption patterns and show that by applying suitable optimization policies directed through our energy consumption models, it is possible to save 20% of energy consumption in cloud data centers. Our research results can be integrated into cloud computing systems to monitor energy consumption and support static and dynamic system level-optimization.
Content may be subject to copyright.
International Journal of Cloud Computing and Services Science (IJ-CLOSER)
Vol.3, No.3, June 2014
ISSN: 2089-3337 31
Journal homepage: IJ-CLOSER
Energy Consumption in Cloud Computing Data Centers
Awada Uchechukwu, Keqiu Li, Yanming Shen
School of Computer Science and Technology
Dalian University of Technology, Dalian, 116024, China
Article Info
Article history:
Received 31 May, 2014
Accepted 6 June, 2014
The implementation of cloud computing has attracted computing as a
utility and enables penetrative applications from scientific, consumer
and business domains. However, this implementation faces
tremendous energy consumption, carbon dioxide emission and
associated costs concerns. With energy consumption becoming key
issue for the operation and maintenance of cloud datacenters, cloud
computing providers are becoming profoundly concerned. In this
paper, we present formulations and solutions for Green Cloud
Environments (GCE) to minimize itsenvironmental impact and energy
consumption under new models by considering static and dynamic
portions of cloud components. Our proposed methodology captures
cloud computing data centers and presents a generic model for them.
To implement this objective, an in-depth knowledge of energy
consumption patterns in cloud environment is necessary. We
investigate energy consumption patterns and show that by applying
suitable optimization policies directed through our energy
consumption models, it is possible to save 20% of energy consumption
in cloud data centers. Our research results can be integrated into cloud
computing systems to monitor energy consumption and support static
and dynamic system level-optimization.
Cloud computing
Green cloud environment
Energy efficiency
Resource management
Energy saving
Copyright © 2014 Institute of Advanced Engineering and Science.
All rights reserved.
Corresponding Author:
Awada Uchechukwu,
Departement of Computer Science and Technology,
Dalian University of Technology,
No.2 Linggong Road, Ganjingzi District, Dalian City, Liaoning Province, P. R. C., 116024.
Recently, the emerging cloud computing offers new computing models where resources such as online
applications, computing power, storage and network infrastructure can be shared as services through the
internet [1]. The popular utility computing model adopted by most cloud computing providers (e.g., Amazon
EC2, Rackspace) is inspiring features for customers whose demand on virtual resources vary with time. The
wide scale potential of online banking, social networking, e-commerce, e-government, information processing
and others,result in workloads of great range and vast scale. Meanwhile, computing and information processing
capacity of several private corporation and public organizations ranging from transportation to banking and
manufacturing to housing have been increasing speedily. Such a vast and vivid increase
ISSN: 2089-3337
IJ-CLOSER Vol. 3, No. 3, June 2014
in the computing resources requires a scalable and efficient information technology (IT) infrastructure
including servers, electrical grids, physical infrastructure, storage, network bandwidth, personnel and huge
capital expenditure and operational cost. Cloud datacenters are the strength of today’s demanding IT
infrastructure, there is crucial need to improve its efficiency.
1.1 Energy-efficient cloud environment
As shown in Fig. 1, cloud computing environment is a large cyber-physical system consisting of
electrical, mechanical and IT systems running a variety of tasks on a multitude of server pools and storage
devices connected with multitier hierarchy of aggregators, routers and switches.
Energy consumption is the key concern in content distribution system and most distributed systems
(Cloud systems). These demand an accumulation of networked computing resources from one or multiple
providers on datacenters extending over the world. This consumption is censorious design parameter in modern
datacenter and cloud computing systems. The power and energy consumed by the compute equipment and the
connected cooling system is a major constituent of these energy cost and high carbon emission.
The energy consumption of date centers worldwide is estimated at 26GW corresponding to about
1.4% of worldwide electrical energy consumption with a growth rate of 12% per year [2], [3]. The Barcelona
medium-size Supercomputing Center ( a data center) pays an annual bill of about £1 million only for its energy
consumption of 1.2 MV [4], which is equivalent to the power of 1, 200 houses [5]. Considering a U.S.
Environmental Protection Agency (EPA) report to Congress [6], in which it is reported that U.S. datacenters
Mechanical Systems
Hot & Cold Alsle
Air Conditioning
In-row Cooling Systems
Electrical Systems
Power Distribution
Uninterruptible Power
Supply (UPS) Systems
Standby Power Systems
IT Systems
Multi-tier Networking
(aggregators, switches,
Computer Infrastructure
( server pools)
Storage Infrastructure
Enterprise Core
Racks, tray, cabinets,
pods, rooms
Fiber uplinks, 1Gb,
10Gb, 40Gb Ethernet
Connectivity &
Security Layer ad Safe
Fallover Services
Load Balancing, Power/
Thermal Management
Middleware – Enterprise
Application Integration,
Data Integration, ORBs
Database Service – MS
SQL, Oracle, Sybase
Web services – Portals,
web-based warehouses
Application Service –
Enterprise Resource
Core Service – DNS,
Fig. 1: High-level components comprising cloud computing environment
IJ-CLOSER ISSN: 2089-3337
Energy Consumption in Cloud Computing Data Centers (Uchechukwu at al.)
Table 1. Cloud Variables Definitions
Cloud utilization
Fraction flow rate
System configuration
Electric current
CRAH heat
Transmission rate
Coefficient of performance
Containment index
Loss coefficient
Component utilization
Flow rate
Load distribution coefficient
Reduction factor
consumed 61 billion kilowatt-hours of power in 2006, which constitutes 1.5% of all power consumed in the
U.S. and represents a cost of $4.5 billions.
Electrical consumption of datacenters in the U.S., which hosts precisely 40% of the world’s cloud
datacenters servers, increased by approximately 40% during the financial breakdown [7], while energy
consumed by servers, cooling, communication, storage, and power distribution equipment (PDU) accounts for
between 1.7% and 2.2% [3]. This increased from 0.8% of U.S. energy consumption in 2000 and 1.5% in 2005
[8]. The environmental impact of cloud datacenters was estimated to be 116.2 million metric tons of CO2
in 2006 [6]. Google datacenter used about 2.26 million MW hours of power to operate in 2010, resulting to
carbon footprint of 1.46 million metric tons of carbon dioxide [9].The inter-government Panel on Climate
change has called for the total reduction of 60%-80% by 2050 to avoid huge environmental damage.
Energy costs are the fastest-rising cost element in the data center portfolio, and yet data center managers are
still not paying sufficient attention to the process of measuring, monitoring and modeling energy use in data
1.2 Energy-inefficient cloud environment
Cloud computing environment comprises thousands to tens of thousands of server machines, working
to render services to the clients [10], [11]. Present servers are far from energy uniformity. Servers consume
80% of the peak power even at 20% utilization [12]. The energy non-uniformity server is a key source to energy
inefficiency in cloud computing environment. Servers are often utilized with between 10% to 50% of their
peak load and servers experience frequent idle times [13]. This means that servers are not working at their
optimal power-performance tradeoff points mostly, and idle mode of servers consumes big portion of overall
power. Another key contributor to power inefficiencies in cloud computing environment is the energy cost of
Cooling and Air Conditioning Units (CACU), accounting to about 30% of the overall energy cost of cloud
environment [14]. These values are reduced by introducing new cooling methods and new server and rack
configurations for cloud computing environments. However, these values can also be reduced drastically for
cloud datacenters located in good geographical locations so that they can benefit from ambient cooling. Yet,
cooling energy consumption in cloud datacenters is still a major contributor to energy inefficiencies in cloud
computing environments.
Yet another reason for energy inefficiency in cloud datacenters is the need for multiple power
conversions in the power distribution system. Precisely, the main ac supply from the grid is first connected to
dc so that it can be used to charge the battery backup system. The output of this electrical energy backup system
then goes through an inverter to produce as power, which is then distributed throughout the cloud environment.
These conversions are necessary due to the oversized and highly redundant uninterrupted power supply (UPS)
modules, which are deployed for voltage regulation and power backup in cloud computing environment.
However, most UPS modules in cloud datacenters operate at 10%-40% of their full capacity [15].
Unfortunately, the UPS conversion efficiency is quite low.
The power usage effectiveness (PUE), which explains how much power is lost in power distribution and
conversion as well as in cooling and air conditioning in cloud computing environments, is calculated as the
ISSN: 2089-3337
IJ-CLOSER Vol. 3, No. 3, June 2014
ratio of the total energy consumption in a cloud datacenter to the overall IT equipment power consumption
[16]. The PUE metric has been steadily reducing over the last decade. In 2003, the PUE metric for a typical
datacenter was estimated to be about 2.6 [17]. In 2010, Koomey estimated that the average PUE was between
1.83 and 1.92 [3]. Most recent cloud datacenters built by Google, Microsoft and Facebook have pushed PUEs
under 1.2 or 1.1 [18, 19]. The cloud datacenter energy efficiency (CDCEE) metric may thus de defined as
where the IT utilization (
) denotes the ration of average IT use over the peak IT capacity in he cloud
datacenter, and the IT efficiency (
) is the amount of useful IT work done per joule of energy.
1.3 Improving energy-efficiency in cloud computing environment
It is appropriate to attain energy proportionality at the server pool or cloud datacenter levels by
dynamically shifting tasks among server and doing server consolidations so that the specific shape of the power
dissipation versus utilization curve at the server level becomes less important, while the shape of the power-
utilization curve at the cloud datacenter level becomes a line that goes through the origin [19]. Also, it has
shown that energy-proportional operation can be attained for lightly utilized servers with full-system
coordinated idle low-power modes [33]. Effects of using energy-proportional servers in datacenters are studied
in [16]. The authors reported 50% energy consumption reduction by using energy-proportional servers with
idle power of 10% of peak power instead if typical servers with 50% idle power consumption. The authors
showed that increasing the energy efficiency of the disk, memory, network cards, and CPU helps in creating
energy-proportional servers. Furthermore, dynamic power management (DPM) techniques, such as dynamic
voltage scaling (DVS) and sleep mode for disk and CPU components, improve the energy proportionality of
the servers.
High energy efficiency in cloud computing environments may be achieved by replacing traditional
cloud datacenter equipment with more-powerful and energy-efficient state-of-the-art servers. These servers use
more advanced internal cooling systems with less energy consumed by their fans. This is important because
internal server energy consumption reductions are amplified by savings in the rack and cloud datacenter power
distribution and cooling systems.
System-wide power management is an important key technique for improving energy efficiency in
cloud computing environments. First, there is the total cost of ownership (TCO) for cloud computing
environments, which includes the energy cost of operating a cloud datacenter. To minimize this cost, the cloud
datacenter’s overall power dissipation must be decreased. Secondly, there is the peak capacity of the power
sources for cloud datacenters and electrical current limitations of the power delivery network in the cloud
datacenter, which set a limit on the peak power draw at the server and datacenter levels.
Maximizing cooling efficiency is another way to lower the energy cost of cooling a cloud datacenter
by deploying computer room air conditioning (CRAC) units and air handling units with demand-driven,
variable frequency drive (VFD) fans within heat exchanges so as to match variable heat loads with variable
airflow rates.
Finally, minimizing this energy consumption can result to conceal cost reduction. Apart from the
enormous energy cost, heat released increases with higher power consumption, thus increases the probability
of hardware system failures [20]. Minimizing the energy consumption has a momentous outcome on the total
productivity, reliability and availability of the system. Therefore, minimizing this energy consumption does
not only reduce the huge cost and improves system reliability, but also helps in protecting our natural
environment. Thus, reducing the energy consumption of cloud computing system and data center is a challenge
because data and computing application are growing in a rapid state that increasingly disks and larger servers
are required to process them fast within the required period of time.
1.4 Paper overview and outline
To deal with this problem and certifying the future growth of cloud computing and data centers is
maintainable in an energy-efficient manner, particularly with cloud resources to satisfy Quality of Service
(QoS) requirement specified by users via Service Level Agreements (SLAs), reducing energy consumption is
IJ-CLOSER ISSN: 2089-3337
Energy Consumption in Cloud Computing Data Centers (Uchechukwu at al.)
necessary. The main objective of this work is to present a new energy consumption models that gives detailed
description on energy consumption in virtualized data centers so that cloud computing can be more
environmental friendly and sustainable technology to drive scientific, commercial and technological
advancements for the future.
The rest of the paper is organized as follows. Section II presents the related work on energy efficiency in
cloud datacenter environments followed by the energy consumption pattern and formulas in Section III. Section
IV formulates the energy consumption models for energy efficiency in cloud computing environments. The
analysis of our energy consumption architecture is defined in Section V. Section VI presents the evaluation
and implementation of our models. Finally, Section VII concludes the paper with discussion on the various
issues and future research directions.
Several issues about green ICT and energy reduction in modern cloud computing systems are receiving
huge attention in the research community. Several other efforts have been made to build energy consumption
models, develop energy-aware cost, manage workload fluctuation and try to achieve an efficient trade-off
between system performance and energy cost. In [21] the authors proposed a cost model for calculating the
total cost of utilization and ownership cost in cloud computing environments. They developed measurable
metrics for this calculation. However, their calculation granularity is based on a single hardware component.
Energy management techniques in cloud environments have also been investigated in the past few years.
In [22] described how servers can be turned ON/OFF using Dynamic Voltage/Frequency Scaling (DVFS)
approach to adjust servers’ power statues. DVFS adjust CPU energy consumption according to the workload.
However, the scope is limited to the CPUs. Therefore, there is a need to look into the behavior of individual
VMs. These can be possible by monitoring the energy profile of individual system components such as CPU,
memory (at run time), disk and cache. Anne at al. [28] observed that nodes still consumed energy even when
they are turned off, due to the card controllers embedded in the nodes which are used to wake up the remote
Sarji et al. [29] proposed two energy models for switching between the server’s operational modes. They
analyze the actual power measurement taken at the server’s AC input, to determine the energy consumed in
the idle state, the sleep state and the off state, to effect switching between this states. However, switching
between power modes takes time and can translate to degraded performance if load goes up unexpectedly.
Moreover, the set of servers that serve the load can also vary continually (a result of load balancing), leading
to short idle times for most servers.
The power modeling techniques has been proposed by several authors. The power consumption model
proposed by Buyya et al. [23] observed a correlation between the CPU energy utilization and the workload
with time. Bohra et al. [24] also proposed a power consumption model that observed a correlation between the
total system’s power consumption and component utilization. The authors created a four-dimensional linear
weighted power model for the total power consumption.
The work done by Chen et al. [25] treats a single task running in a cloud environment as the fundamental
unit for energy profiling. With this technique, Chen and her colleagues observed that the total energy
consumption of two tasks is not equivalent to the sum of individual consumed energy due to scheduling
overhead. They created a power model for total energy consumption which focuses on storage, computation
and communication resources.
On the other hand, several research effort has been also been made to minimize energy consumption in
cloud environments mostly on virtualization. This technology permits one to overcome power clumsy by
accommodating multiple VMs on a single physical host and by performing live migrations to optimize the
utilization of the available resources. Yamini et al. [26] proposed a cloud virtualization as a potential way to
reduce global warming and energy consumption. Their approach utilizes less number of servers instead of
using multiple servers to offer service for multiple devices.
The power modeling techniques for the physical infrastructure (power and cooling systems) in data
centers, proposed by Pelley et al. [34] is most relevant for us. They worked out first models which try to capture
a data center at large.
In this paper, we provide tools to the Cloud computing environments to asses and reason about total
energy consumption. Our approach target both Cloud data center simulation and analytic models. We show
ISSN: 2089-3337
IJ-CLOSER Vol. 3, No. 3, June 2014
that by applying energy optimization policies through energy consumption models, it is possible to save huge
amount of energy in cloud environments and data centers.
The understanding of energy consumption pattern is necessary for its improvement. Servers consume
a larger fraction of energy in cloud environments and their energy consumption varies with utilization. This
consumption also vary with the type of computation going on in the server e.g., data retrieval and data
Networking equipment, lightning and pumps also contribute to total energy consumption. However,
the contribution of each totals a few percent of the overall consumption. Since these systems energy
consumption does not vary significantly with data center load, we account for these systems by as a fixed
energy overheads (around 6%) of the baseline power.
The power conditioning system supplies power to the uninterrupted power supplies (UPSs). The UPSs charge
continuously and supply power until generators can start during a utility failure. UPS distribute electricity at
high voltage (480V-1KV) to power distribution units (PDUs), which regulate voltage to match IT equipment
requirements. The PDUs and UPSs consume significant amount of energy, and their consumption increases
with workload.
The cooling system maintains humidity and air quality while it evacuates heat from the facility. This
heat is the result of power dissipation. Removing this heat while maintaining humidity and air quality requires
an extensive cooling system. Cooling starts with the computer room air handler (CRAH), which transfer heat
servers’ hot exhaust to a chilled water cooling loop while supplying cold air all over the facility. Extracting
heat in this manner requires huge amount of energy. CRAH unit energy consumption dominates the total
cooling system energy consumption and its requirement increases with both cloud environment thermal load
and outside temperature. These overheads can be approximately modeled as a fixed figure of total energy
consumed [30]. The power and cooling flow is presented in Fig. 2 and marketed by different industry segments
which account for most of data center’s energy consumption: (1) servers and storage systems, (2) power
conditioning equipments, (3) cooling and humidification systems, (4) networking equipments, and (5) lighting
/physical security. Thus, the first three sub-systems dominate power draw in cloud environments and data
centers when in active mode.
However, the distribution of load in each sub-system can affect energy consumption, because of the
non-linear interaction between sub-systems. Several research studies have proposed server energy consumption
[8][9][10]. Whereas, the actual amount of utilization-energy varies, servers generally consume roughly half of
their peak load energy when in idle mode, and energy consumption increases with resource utilization. Energy
consumed during idle mode is a fixed part of the overall consumption. The dynamic part of energy consumption
is the additional energy consumed by running tasks in the Cloud. Therefore, we divide energy consumption
into two parts:
Fixed energy consumption (energy consumed during server idle state)
Dynamic energy consumption (energy consumed by Cloud tasks and cooling system)
Fig. 2: Power and cooling flow of cloud environment
IJ-CLOSER ISSN: 2089-3337
Energy Consumption in Cloud Computing Data Centers (Uchechukwu at al.)
3.1 Energy consumption formulas
The total energy consumption of an active server for a given time frame is the sum of energy consumed
when the system is fixed and dynamic defined as
is formulated as follows:
Total Fix DynE E E    
There is additional energy is generated by scheduling overhead, denoted by
. This makes the energy
consumption of two tasks, not equivalent to the sum of individual consumed energy. In this paper, we focus
on the energy consumed by, server idle state, cooling systems, computation, storage and communication
resource utilizations. These are defined as follows:
1. Energy consumption of server idle mode is denoted by
2. Energy consumption of cooling system is denoted by
3. Energy consumption of computation resources is denoted by
4. Energy consumption of storage resources is denoted by
5. Energy consumption of communication resources is denoted by
Therefore, the above formula (1) can be transformed into:
(Total Idle Cool CommuE E E E  
)Store Compu SchedE E E 
As discussed in the requirements, time variations in renewable energy availability and cloud computing
environments efficiencies provide both opportunities and challenges for managing IT workload and cloud
computing datacenters. In this section, we present novel models for energy efficiency aware management to
improve the sustainability of cloud computing datacenters. In particular, we formulate measurable metrics
based on runtime tasks to compare rationally the relation existing between energy consumption and cloud
workload and computational tasks, as well as system performance.
Our models relate to overall cloud computing environments energy consumption to total utilization,
represented by
. We represent the per-server utilization with
, where the subscript denotes the server in
question as follows:
[ ]
Srv Srv i
ISSN: 2089-3337
IJ-CLOSER Vol. 3, No. 3, June 2014
Table 2: Value of
for Intel Processors
Processor type
Intel Xeon dual-core E5502
Intel Xeon quad-core E5540
Intel Xeon hexa-core X5650
We characterize the individual server utilization,
, as a function of
and a measure of task consolidation,
, to abstract the effect of consolidation.
is used to capture the degree of which load is distributed across
cloud datacenter’s servers. We define individual server utilization as:
(1 )
Srv U
 
Only holds meaning for the
( (1 ) )SrvN U U  
servers that are non-idle. Fig. 3 depicts the relationship
, and
= 0 corresponds to perfect consolidation. The cloud’s workload is packed onto the
minimum number of servers, and the utilization of any active server is 1.
= 1 represents the opposite extreme
of perfect load balancing. All servers are active with
4.1 Modeling server idle state energy consumption
The idle energy consumption of a server can be determined by applying the following well known
equation [38] from Joule’s Ohm’s law:
E I V 
denotes the energy or power (Watt),
represents the electric current (Ampere) and
indicates the
voltage. The above equation can be adopted in order to determine the idle energy consumption at core level by
assuming that each core contributes equally to the overall idle energy consumption of a processor:
i i iE I V 
represents respectively the power, current and voltage of the corresponding core
Furthermore, by analyzing the current
and voltage
relationship, we derive the following second order
polynomial to model the current leakage as follows:
i iI V V   
= 0.114312 ((
( )V
= 0.22835
( )
= 0.139204 (
) are the coefficients [39].
With the implementation of energy saving mechanisms (e.g. AMD’s PowerNow! and Cool ‘n’ Quite, Intel’s
SpeedStep), the idle energy consumption of a core (processor) decreases significantly. This is actualized by
decreasing the Dynamic Voltage and Frequency Scaling (DVFS) of a core.
IJ-CLOSER ISSN: 2089-3337
Energy Consumption in Cloud Computing Data Centers (Uchechukwu at al.)
r i iE E
is the factor for reduction in the power consumption
of core
, whereas
is the reduced power
consumption of core
. However,
vary depending on the energy saving mechanism in use.
Given a processor of
numbers of cores with a specific energy saving mechanism, then its idle energy
consumption is given by:
irE E
The values of the reduction factor
for different types of Intel processors
4.2 Modeling cooling systems energy consumption
The CRAH unit energy consumption dominates the total cooling system energy consumption, it
transfer heat servers’ hot exhaust to a chilled water cooling loop while supplying cold air all over the facility.
Its requirement increases with both cloud environment thermal load and outside temperature. The efficiency
of cooling process varies on the speed of the air exiting the CRAH unit, the substance used in the chiller, etc.
In general, heat is transferred between two bodies according to the thermodynamic principle as follows:
( )hc ha caC T T 
is the power transferred between a device and fluid
represents the fluid mass flow, and
is the
specific heat capacity of the fluid.
represent the hot and cold temperatures respectively. The value
depend on the physical air flow throughout the data center and air recirculation.
Cloud datacenter designers use computational fluid dynamics (CFD) to model the complex flow and
CRAHs to minimized recirculation. We replace CFD with simple parametric model that capture its effect on
cloud computing energy consumption. Based on previous metrics for recirculation [35], [36], we introduce a
containment index (
). Containment index is defined as the fraction of air ingested by a server that is
Fig. 3: Individual server utilization
0 20 40 60 80 100
Data Center Utilization (%)
Server Utilization (%)
load distribution = 0.02
load distribution = 0.25
load distribution = 0.5
load distribution = 0.75
load distribution = 1
Fig. 4: CRAH supply temperature
0.9 0.92 0.94 0.96 0 .98 1
Containment Index
CRAH Supply Temperature (
Data Center Utilization = 10%
Data Center Utilization = 40%
Data Center Utilization = 70%
Data Center Utilization = 100%
ISSN: 2089-3337
IJ-CLOSER Vol. 3, No. 3, June 2014
supplied by a CRAH. Thus, a
of 1 implies no recirculation from the device. Our model uses a single, global
containment index to represent average behavior, resulting as follows:
( )
air ha ca
air hc a aC T T  
is the heat transferred by the server or CRAH,
represents the total air flowing through the
the temperature of the air exhausted by the server, and
is the temperature of the cold air
supplied by the CRAH. CRAHs transfers heat out of the server room to the chilled water loop. Thus, we model
equation (12) above using a modified effectiveness-NTU method [37]:
0.7( (1 ) )CRAH air ha ca
air hc a a wtE C f T T T    
is the heat removed by the CRAH.
is the transfer efficiency at the maximum mass flow rate (0 to 1),
represents the volume flow rate, and
the chilled water temperature.
The efficiency of a CRAH unit is measured using the Coefficient of Performance (COP), which is defined as
the ratio of the amount of heat that is removed by the CRAH unit (
) to the total amount of energy that is
consumed in the CRAH unit to chill the air (
) [31]:
The COP of a CRAH unit varies by the temperature (
) of the cold air that it supplies to the cloud
facility. The summation of energy consumed by the CRAH (
) and IT (
) equipments in cloud
environment equal the total power dissipation [32]. The energy consumed by the CRAH unit may be specified
Fig. 5: CRAH power
0 0.2 0.4 0.6 0.8 1
Server Utilization
CRAH Power (kW)
Chilled Water = 52oF
Chilled Water = 49oF
Chilled Water = 46oF
Chilled Water = 43oF
Chilled Water = 40oF
Fig. 6: Chilled Water Temperature
35 40 45 50 5 5 60
Chilled Water Temp (oF)
Chiller Power (kW)
IJ-CLOSER ISSN: 2089-3337
Energy Consumption in Cloud Computing Data Centers (Uchechukwu at al.)
( )
Energy consumed by the CRAH unit is dominated by fan power, which increases dynamically with
the cube of mass flow rate (
) to some maximum amount. Additionally, some fixed energy is consumed by
sensors and control system. Thus, the energy consumed by the CRAH unit totals it’s fixed and dynamic activity:
Idle Dyn
The efficient of heat exchange and the mass available to transfer heat increases as the volume flow
rate through the CRAH increases. We use a single value of
to simplify our model by allowing the
conservation of air flow between the CRAH and servers. Fig. 4 demonstrates air recirculation places on the
cooling system. The figure shows the CRAH supply temperature for typical maximum safe server inlet
temperature of 770F. As
decreases, the required CRAH supply temperature quickly drops. Thus, lowering
supply temperature results in superlinear increases in CRAH and chiller plant power and preventing are
recirculation can drastically improve cooling efficiency.
The effects of containment index and chilled water supply temperature on CRAH power are shown in
Fig. 5. Here the CRAH model has a peak heat transfer efficiency of 0.5, a maximum airflow of 6900 CFM,
peak fan power of 3KW, and idle energy consumption cost of 0.1KW. When the chilled water supply
temperature is low, CRAH units are relatively insensitive to changes in containment index. For this reason,
cloud computing operators often chosse low chilled water supply temperature, leading to overprovisioned
cooling in the common case.
Chillers at a constant outside temperature and chilled water supply temperature will require energy
that grows quadratically with the quantity of heat removed with utilization. The HVAC community has
developed several modeling approaches to assess chiller performance. Although physics-based models do
exist, we chose the Department of Energy’s DOE2 chiller model [41]. Fitting the DOE2 model to a particular
chiller requires numerous physical measurements. We use a benchmark set of regression curves provided by
the California Energy Commission [42].
A chiller intended to remove 8MW of heat at peak load using 3,200 KW at a steady outside air
temperature of 850F, a steady chiller water supply temperature of 450F, and a cloud load balancing coefficient
of 0 will consume the following power as a function of total cloud utilization (KW):
742.8 1,844.6 538.7ChillerE U U 
Fig. 6 demonstrates the energy required to supply successively lower chilled water temperature at
for an 8MW peak thermal load. However, as thernal load increases, the energy required to lower
the chilled water temperature becomes substantial. The difference in chilled power for a 450F and 550F chilled
water supply at peak load is nearly 500KW. Fig. 7 displays the rapidly growing energy reqiurement as cooling
load increases for a 450F chilled water supply. The grah also shows the strong sensitivity of chiller energy
consumption to outside air temperature.
ISSN: 2089-3337
IJ-CLOSER Vol. 3, No. 3, June 2014
4.3 Modeling power conditioning systems energy Consumption
Cloud computing environments need considerable infrastructure simply to supply uninterrupted,
stable electric power. Power distribution units transform the high voltage power supplied throughout the cloud
environment to voltage levels appropriate for servers. They incur a constant energy loss as well as a energy
loss proportional to the square of the load [40]:
Idle Servers
represents the energy consumed by the PDU,
denotes the PDU energy loss coefficient, and
the PDU’s idle energy consumption. UPSs provide temporary energy supply during utility failure.
UPS systems are placed in series between the utility supply and PDUs and impose some energy consumption
overhead even when operation on utility energy. UPS energy overheads follow the relation [40]:
denotes the UPS loss coefficient. UPS losses about 9% of their input energy at full load.
Fig. 8 shows the power losses for a 10MW of cloud environment. At peak load, power conditioning
loss is 12% of total server energy consumption. These losses result in additional heat that must be evacuated
by the cooling system.
4.4 Modeling dynamic derver state energy consumption
The energy consumption of a task (communication, storage, computation) is determined by the
number of processes, size of data to be processed, size of data to be transmitted and the system configuration (
). The energy consumption profiling metrics are presented in Table 3. Thus, the energy consumed by each
task can be formulated as:
Fig. 7: Effects of
on P Chiller
0 20 40 60 80 100
Data Center Utilization (%)
Chiller Power (kW)
TOutside = 95oF
TOutside = 85oF
TOutside = 75oF
TOutside = 65oF
TOutside = 55oF
Fig. 8: Power conditioning losses
0 20 40 60 80 100
Data Center Utilization (%)
Power Los s (kW)
Power Conditioning Loss
UPS Loss
PDU Loss
IJ-CLOSER ISSN: 2089-3337
Energy Consumption in Cloud Computing Data Centers (Uchechukwu at al.)
( , , , )
i i i i i i
Em m Nm Dm Tm SCf
( , , , )S S S S S i
i i i i i
E f N D T SC
( , , , )
i i i i i i
Ec c Nc Dc Tc SCf
The energy consumed by cloud tasks
is formulated as follows:
1 1 1
n i
i i i
i i i
DynE Em Es Ec
 
 
 
Adding the energy generated by schedule overhead and interference, equation (8) above can be transformed as
1 1 1 Sched
n i
i i i
i i i
DynE Em Es Ec E
 
  
 
The architecture of our energy-saving mechanism, presented in Fig. 9, is based on Optimization,
Reconfiguration and Monitoring. The entire state of Cloud environment is automatic monitored. Another major
contribution of this paper is committed to detailed analysis of the state of Cloud environments and data centers
resources with relevant energy consumption attributes and interconnections.
This state is recurrently analyzed by the Optimization module in order to find a surrogate software
application and service allocated configurations that enables energy minimization. Once an appropriate energy-
saving configuration is detected, the loop is closed by issuing a set of action on Cloud environment to
reconfigure the allocation of this energy-saving setup.
Monitoring and Reconfiguration modules communicate with the Cloud environment monitoring
framework to perform their tasks. The Optimization module ranks the target configurations, this is established
by applying energy-saving policies without violating existing SLAs, with respect to their energy consumption
that are predicted by the Energy Calculator module. The accuracy predictions of this module is essential to take
the most appropriate energy minimization decisions, it has the ability to forecast the energy consumption of
Cloud environment after a possible reconfiguration option.
Energy Consumed
Process Number
Size of Data Processed
Size of Data Transmitted
ISSN: 2089-3337
IJ-CLOSER Vol. 3, No. 3, June 2014
In this section, we demonstrate the utility of our models. We provide a comparison of power
requirement between several presumptive Cloud data centers using the energy consumption models [34, 43].
Each scenario based on the previous, present new power-saving attribute. We decompose each data center on
25% utilization. Next, we present the configured Cloud data centers as well as induced workload.
6.1 Scenario 1 and 2
These represent conventional Cloud data center with legacy physical infrastructure typical of facilities
commissioned in the last three to five years. We use yearly average for outside air temperature (oF). We assume
limited server consolidation and a relatively poor containment index of 0.9. Furthermore, we assume typical
(inefficient) servers with idle power at 60% of peak power, and static chilled water and CRAH air supply
temperature set to 45oF and 65oF, respectively. We scale the Cloud date center such that the Scenario 1 facility
consumes precisely 10MV at peak utilization.
6.2 Scenario 3
This represents a data center where virtual machine consolidation or other mechanisms reduce
from 0.26 to 0.57 and reducing the number of active servers from 81% to 44%. Improved consolidation
drastically decreases aggregate power draw, but, paradoxically, it increases PUE. These results illustrate the
shortcoming of PUE as a metric for energy efficiency; it fails to account for the inefficiency of IT equipment.
6.3 Scenario 4
This allows servers to idle at 5% of peak power by transitioning rapidly to a low power sleep state,
reducing overall data center power by another 22% [33]. This approach and virtual machine consolidation take
alternative approaches to target the same source of energy inefficiency: server idle power waste.
6.4 Scenario 5
This posits integrated, dynamic control of the cooling infrastructure. We assume an optimizer with
global knowledge of data center load/environmental conditions that seek to minimize chiller power. The
optimizer chooses the highest
that still allows CRAHs to meet the maximum allowable server inlet
temperature. This scenario demonstrates the potential for intelligent cooling management.
Fig. 9: Energy consumption architecture
IJ-CLOSER ISSN: 2089-3337
Energy Consumption in Cloud Computing Data Centers (Uchechukwu at al.)
Table 1: Presumptive cloud data centers
6.5 Scenario 6
Finally, this represents a data center with a containment system (e.g., servers enclosed in shipping
containers), where containment index is increased to 0.99. Under this scenario, the cooling system power draw
is drastically reduced and power conditioning infrastructure becomes the limiting factor on power efficiency.
We have presented holistic models of Cloud data center fundamentals reasonable to use in a detailed Cloud
environment simulation infrastructure, abstract estimation and green energy prediction as an effective solution.
Figure 10 displays the respective energy draws of our scenarios, actualizing 30% better energy efficiency and
the list of scenarios is shown on Table 4.
Cloud computing is becoming more and more crucial in IT sector due to abundant advantages it
renders to its end users. With the high user demands, Cloud environment possess very large ICT resources. To
this, power and energy consumption of Cloud environment have become an issue due to ecological and
economical reasons. In this paper, we have presented energy consumption formulas for calculating the total
energy consumption in Cloud environments and show that there are incentives to save energy. To this respect,
we described an energy consumption tools and empirical analysis approaches. Furthermore, we provide generic
energy consumption models for server idle and server active states. This research result is crucial for
developing potential energy legislation and management mechanisms to minimize energy consumption while
system performance is achieved for Cloud environments.
Idle PeakE E
( )T F
Fig. 10: Power saving features
1 2 3 4 5 6
Total Power (KW)
ISSN: 2089-3337
IJ-CLOSER Vol. 3, No. 3, June 2014
As future work, we will investigate several Cloud environments and propose new optimization
policies which will minimize the CO2emissions of Cloud environment, we will integrate energy cost rate into
our new models in differing environmental impact and to minimize the total energy cost.
This work is supported by the National Science Foundation for Distinguished Young Scholars of China
(Grant No. 61225010), NSFC under grant nos.61173160, 61173161, 61173162, 61173165 and New Century
Excellent Talents in University (NCET-10-0095) of Ministry of Education of China.
[1] M. Armbrust, “Above the clouds: A Berkeley view of cloud computing”, Technical Report UCB/EECS -2009-
28, 2009.
[2] BONE project, “WP 21 Tropical Project Green Optical Networks: Report on year 1 and unpdate plan for
activities”, No. FP7-ICT-2007-1216863 BONE project, Dec. 2009.
[3] J. Koomey, “Estimating Total Power Consumption by Server in the U.S and the World”, February 2007,
[4] Jordi Toress, “Green Computing: The next wave in computing”, In Ed. UPC Technical University of Catalonia,
February 2010.
[5] Peter Kogge,”The Tops in Flops”, pp. 49-54, IEEE Spectrum, February 2011.
[6] U.S Environmental Protection Agency. (2006). Report to Congress on Server and Datacenter Energy Efficiency
Public Law [Online]. Available:
[7] Greenpeace. (2011). Greenpeace “Likes” Facebooks New Datacenter, But Wants aGreener Friendship[Online].
Available: http://www. books-new-
[8] J. Kovar. (2011, Aug. 10). Data Center Power Consumption Grows Less Than Expected: Report
[Online]. Available: http://www. grows-
[9] R. Miller. (2011, Sep.). Googles Energy Story: High Efficiency, Huge Scale [Online]. Available: archives/2011/09/08/googles-energy-story-high-efficiency-huge-scale
[10] M. Armbrust, A. Fox, R. Griffith, A. D. Joseph, R. Katz, A. Konwinski, G. Lee, D. Patterson, A. Rabkin, I.
Stoica, and M. Zaharia, A view of cloud computing,Commun. ACM, vol. 53, no. 4, pp. 5058, 2010.
[11] R. Buyya, “Market-oriented cloudcomputing: Vision,hype, andreality of delivering computing as the5th utility,
in Proc. Int. Symp. Cluster Comput. Grid, May 2009, p. 1.
[12] L. A. Barroso and U. Hölzle, “The case for energy-proportional computing,IEEE Comput., vol. 40, no. 12, pp.
3337, Dec. 2007.
[13] L. A. Barroso and U. Hölzle, "The Datacenter as a Computer: An Introduction to the Design of Warehouse-
Scale Machines". San Rafael, CA: Morgan and Claypool, 2009.
[14] N. Rasmussen, “Calculating total cooling requirements for
Amer. Power Convers., white paper 25,
[15] Data Center Energy Efficiency Training, U.S. Department of
(2011). Electrical Systems [Online].
[16] C. Belady, A. Rawson, J. Pfleuger, and T. Cader. Green Grid Datacenter Power Efficiency
Metrics: PUE and DCIE [Online]. Available:
DataCenterPowerEfficiencyMetrics PUE
[17] S. Ghemawat, H. Gobioff, and S.-T. Leung, “The Google file
in Proc. ACM Symp. Operating Syst.
Principles, 2003, pp. 2943.
[18] S. Higginbotham. (2012, M ar .). Whose D at a C en te rs Ar e M or e Efficient? Facebooks or Googles?
[Online]. Available: http:// or-googles/
[19] T. Seubert. (2012, Feb. 27). Microsoft Builds New Data Center
Dublin [Online]. Available:
[20] W.C. Feng, X. Feng and C. Rong, "Green Supercomputing Comes of Age," IT Professional, vol.10, no.1, pp.17-
23, Jan.-Feb. 2008.
[21] X. Li, Y. Li, T. Liu, J. Qiu, and F. Wang, "The method and tool of cost analysis for cloud computing," in the
IEEE International Conference on Cloud Computing (CLOUD 2009), Bangalore, India, 2009, pp. 93-100.
[22] L. Shang, L.S. Peh, and N. K. Jha, "Dynamic voltage scaling with links for power optimization of interconnection
networks," In the 9th International Symposium on High-Performance Computer Architecture (HPCA 2003),
Anaheim, California, USA, 2003, pp. 91-102.
[23] B. Rajkumar, B. Anton and A. Jemal, “Energy Efficient Management of Data Center Resources for Cloud
Computing: A Vision Architectural Elements and Open Challenges”, In Proc. International Conference on
IJ-CLOSER ISSN: 2089-3337
Energy Consumption in Cloud Computing Data Centers (Uchechukwu at al.)
Parallel and Distributed Processing Techniques and Applications (PDPTA 2010), Las Vegas, USA, July 12-15,
[24] A. Bohra and V. Chaudhary, “VMeter: Power modelling for virtualized clouds," International Symposium
on Parallel & Distributed Processing, Workshops and Phd Forum (IPDPSW), 2010 IEEE, vol., no., pp.1-8, 19-
23 April 2010.
[25] F. Chen, J. Schneider, Y. Yang, J. Grundy and Q. He, “An energy consumption model and analysis tool for Cloud
computing environments," Green and Sustainable Software (GREENS), 2012 First International Workshop on,
vol., no., pp.45-50, 3-3 June 2012.
[26] B. Yamini and D.V. Selvi, "Cloud virtualization: A potential way to reduce global warming," Recent Advances
in Space Technology Services and Climate Change (RSTSCC), 2010, vol., no., pp.55-57, 13-15 Nov. 2010.
[27] Z. Zhang and S. Fu, "Characterizing Power and Energy Usage in Cloud Computing Systems," Cloud Computing
Technology and Science (CloudCom), 2011 IEEE Third International Conference on, vol., no., pp.146-153, Nov.
29 2011-Dec. 1 2011.
[28] A.C. Orgerie, L. Lefevre, and J.P. Gelas, "Demystifying energy consumption in Grids and Clouds," Green
Computing Conference, 2010 International, vol., no., pp.335-342, 15-18 Aug. 2010.
[29] I. Sarji, C. Ghali, A. Chehab, A. Kayssi, "CloudESE: Energy efficiency model for cloud computing
environments," Energy Aware Computing (ICEAC), 2011 International Conference on, vol., no., pp.1-6, Nov.
30 2011-Dec. 2 2011.
[30] X. Fan, W.D. Weber, and L. A. Barroso, "Power provisioning for a warehouse-sized computer," in the 34th
International Symposium on Computer Architecture (ISCA 2007), San Diego, California, USA, 2007, pp. 13-23.
[31] J. Moore, J. Chase, P. Ranganathan, R. Sharma, Making Scheduling "Cool": Temperature-Aware Workload
Placement in Data Centers” In Proc. of the 2005 USENIX Annual Technical Conference, Anaheim, CA, USA,
April 1015, 2005.
[32] P. Ehsan and P. Massoud, “Minimizing data center cooling and server power costs”, In Proc. Of the 4th
ACM/IEEE International Symposium on Low Power Electronic and Design (ISLPED), 2009, Pages 145-150
[33] D. Meisner, B. T. Gold, T. F. Wenisch, “PowerNap: Eliminating Server Idle Power”, In Proc. of the 14th
international conference on Architectural support for programming languages and operating systems (ASPLOS),
USA, 2009.
[34] S. Pelley, D. Meisner, T. F. Wenisch, and J. W. VanGilder, “Understanding and absracting total data center
power,” in WEED: Workshop on Energy Efficienct Design, 2009.
[35] R. Tozer, C. Kurkjian, and M. Salim, “Air management metrics in data centers,” in ASHRAE 2009, January 2009.
[36] J. W. VanGilder and S. K. Shrivastava, “Capture index: An airflow-based rack cooling performance metric,”
ASHRAE Transactions, vol. 113, no. 1, 2007.
[37] Y. A. Çengel, Heat transfer: a practical approach, 2nd ed. McGraw Hill Professional, 2003.
[38] Meade RL, Diffenderfer R (2003) Foundations of Electronics: Circuits & Devices. Clifton Park, New York.
ISBN: 0-7668-4026-3
[39] ZES ZIMMER PA, Brodersen RW (1995) Minimizing Power Consumption in CMOS Circuits. Tech. rep.,
University of California at Berkely.\\Min\_pwr\_consum\_CMOS\_crct/
[40] N. Rasmussen, “Electrical efficiency modeling for data centers,” APC by Schneider Electric, Tech. Rep. #113,
[41] DOE (Department of Energy), “Doe 2 reference manual, part 1, version 2.1,1980.
[42] CEC (California Energy Commission), “The nonresidential alternativecalculation method(acm) approvalmanual
for the compliance with californias 2001 energyefficiency standards,April 2001.
[43] A. Uchechukwu, K. Li and Y. Shen,”Improving Cloud Computing Energy Efficiency.” In Proc. Asia Pacific
cloud Computing Congress, Nov., 2012.
ISSN: 2089-3337
IJ-CLOSER Vol. 3, No. 3, June 2014
Awada Uchechukwu received the B.Sc. degree in computer science from Ebonyi
State University, Abakaliki, Nigeria and the M.Eng degree in computer applied
technology from Harbin Engineering University, Harbin, China, in 2011, he is
currently working towards the Ph.D. degree in computer science and engineering
with the Network and Cloud Computing Laboratory at Dalian University of
Technology. His research interests include energy consumption, cloud computing,
big data and distributed computing. He is a student member of the IEEE Computer
Keqiu Li received B.S. and M.S degree in applied mathematics from Dalain
University of Technology, Dalian, China in 1994 and 1997 respectively and Ph.D.
degree in information technology from Japan Institute of Science and Technology in
2005. Currently, he is a Professor at Dalian University of Technology, where he is
the Dean of the school of computer science and technology and Director of the
network and cloud computing laboratory. His research interests include web
technology, grid/cloud computing, mobile computing, network and security. He is a
senior member of the IEEE and the IEEE Computer Society.
Yanming Shen received the B.S. degree in automation from Tsinghua University,
Beijing, China, in 2000 and the Ph.D. degree from the Department of Electrical and
Computer Engineering at the Polytechnic University (now Polytechnic Institute of
New York University), Brooklyn. He is an Associate Professor in the Computer
Science and Engineering Department at Dalian University of Technology, Dalian,
China. He was a summer intern with Avaya Labs in 2006, conducting research on
IP telephony. His general research interests include cloud computing, distributed
systems, packet switch design, peer-to-peer video streaming, and algorithm design,
analysis, and optimization.
... Many researchers [3,5,11,13,20,23,[25][26][27] have developed resource allocation algorithms for cloud computing. These algorithms are developed based on different cloud environments and objectives to be achieved. ...
... Gamsiz et al. [3] have introduced a combinatorial auction-based model by considering energy in order to solve the cloud resource allocation problem. Uchechukwu et al. [26] have investigated various energy consumption patterns and suggested energy optimization techniques. Wang et al. [27] have introduced a VM placement algorithm, which is based on the energy and quality of service. ...
... [53][54][55][56] proposed an approach that reduces energy consumption by disabling idle virtual machines (VMs) and categorizing active VMs into different energy clusters using a fuzzy C-means algorithm. Consequently, the extent of the Service Level Agreement (SLA) violation was minor [57][58][59][60][61][62][63][64][65][66][67][68]. ...
... Three main vital factors influence energy performance [58]. First, IT manufacturers must continue to improve technology for energy efficiency in IT devices, especially servers and storage drives [59][60][61][62][63][64]. Second, increasing virtualization software deployment, which empowers many applications to run on a single server, undoubtedly reduces every application's energy consumption. ...
Full-text available
Global warming is one of the most compelling environmental threats today, as the rise in energy consumption and CO2 emission caused a dreadful impact on our environment. The data centers, computing devices, network equipment, etc., consume vast amounts of energy that the thermal power plants mainly generate. Primarily fossil fuels like coal and oils are used for energy generation in these power plants that induce various environmental problems such as global warming ozone layer depletion, which can even become the cause of premature deaths of living beings. The recent research trend has shifted towards optimizing energy consumption and green fields since the world recognized the importance of these concepts. This paper aims to conduct a complete systematic mapping analysis on the impact of high energy consumption in cloud data centers and its effect on the environment. To answer the research questions identified in this paper, one hundred nineteen primary studies published until February 2022 were considered and further categorized. Some new developments in green cloud computing and the taxonomy of various energy efficiency techniques used in data centers have also been discussed. It includes techniques like VM Virtualization and Consolidation, Power-aware, Bio-inspired methods, Thermal-management techniques, and an effort to evaluate the cloud data center’s role in reducing energy consumption and CO2 footprints. Most of the researchers proposed software level techniques as with these techniques, massive infrastructures are not required as compared with hardware techniques, and it is less prone to failure and faults. Also, we disclose some dominant problems and provide suggestions for future enhancements in green computing.
... Energy consumed by computing resources and connected cooling facilities is the main component of energy costs and high carbon emissions. According to research conducted by Uchechukwu and Shen [15], it is estimated that energy consumption by data centers around the world is about 1.4% of electricity consumption worldwide and is growing at a rate of 12% annually. The energy consumption of processing units is approximately 42%, cooling facilities about 15.4%, and storage facilities nearly 14.3% [16]. ...
... Step 2) The process of finding the optimal solution starts based on SOA. In this step, based on position information, the amount of cost, makespan, load, energy consumption, and waiting time are calculated according to Equations (14), (15), (18), (23), and (24), respectively. Then, according to Equation (30), the objective function value of each seagull is calculated. ...
Full-text available
Cloud computingprovides computingresources like softwareandhardware as a service by the network for several users. Task scheduling is one of the main problems to attain cost-effective execution. The main purpose of task scheduling is to allocate tasks to resources so that it can optimize one or more criteria. Since theproblemof taskschedulingis oneof the NondeterministicPolynomial-time (NP)-hard problems, meta-heuristicalgorithms have been widely employedforsolvingtask schedulingproblems. One of the new bio-inspired meta-algorithms is Seagull Optimization Algorithm (SOA). In this paper, we proposedan energy-aware andcost-efficient SOA-basedTaskScheduling(SOATS) algorithm. The aims of proposed algorithm to make a trade-off between five objectives (i.e., energy consumption, makespan,cost,waitingtime,andloadbalancing) using a fewer number of iterations. The experiment results by comparing with several meta-heuristic algorithms (i.e., Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), and Whale Optimization Algorithm (WOA)) prove that the proposed technique performs better in solving task scheduling problem. Moreover, we comparedthe proposedalgorithmwith well-known schedulingmethods: Cost-basedJob Scheduling (CJS), Moth Search Algorithm based Differential Evolution (MSDE), and Fuzzy-GA (FUGE). In the heavilyloadedenvironment, the SOATSalgorithmimprovedenergy consumption and cost saving by 10 and 25%, respectively. © 2022 Materials and Energy Research Center. All rights reserved.
... Chen & Long [19] proposed a hybrid approach based on particle swarm optimization, and ant colony optimization techniques. Chechukwu et al., presented a solution for green cloud environment with minimum power consumption and minimum impact of technology on environment [21]. Authors focused on minimization of energy and support the system level optimization. ...
Full-text available
Resources are offered to customers on demand in the modern era of computing, communication, and technology. User demand of the resources depends on service provider and consumer. The optimal assignment of the cloud resources depends on fitness function and resource management technique. In this manuscript, the key focus is to proposed a model based on meta-heuristic evaluation technique. The meta-heuristic evaluation technique provides an optimal placement of the virtual machines to the user requests across the globe. The presented framework, elephant heard optimization with neural network (EHO-ANN) outperforms the existing static, dynamic and nature inspired techniques. The EHO-ANN is evaluated and analyzed against Max-Min, Genetic Approach and BB-BC cost aware approach. The evaluation and analysis include the performance metric, average execution time(ms), and cost.
... In [27], Uchechukwu et al. investigated energy consumption needs in cloud data centers. The authors provided formulas for calculating the total energy consumption of cloud environments, validating the models in six scenarios shown as examples. ...
Full-text available
The digital transition that drives the new industrial revolution is largely driven by the application of intelligence and data. This boost leads to an increase in energy consumption, much of it associated with computing in data centers. This fact clashes with the growing need to save and improve energy efficiency and requires a more optimized use of resources. The deployment of new services in edge and cloud computing, virtualization, and software-defined networks requires a better understanding of consumption patterns aimed at more efficient and sustainable models and a reduction in carbon footprints. These patterns are suitable to be exploited by machine, deep, and reinforced learning techniques in pursuit of energy consumption optimization, which can ideally improve the energy efficiency of data centers and big computing servers providing these kinds of services. For the application of these techniques, it is essential to investigate data collection processes to create initial information points. Datasets also need to be created to analyze how to diagnose systems and sort out new ways of optimization. This work describes a data collection methodology used to create datasets that collect consumption data from a real-world work environment dedicated to data centers, server farms, or similar architectures. Specifically, it covers the entire process of energy stimuli generation, data extraction, and data preprocessing. The evaluation and reproduction of this method is offered to the scientific community through an online repository created for this work, which hosts all the code available for its download.
... For example, according to the article of Rong et al. (2016: 689) investigating optimizing energy consumption for data centers, energy saving technologies that can be used for data centers provide data centers with significant advantages in terms of energy savings. Likewise, the results of the research conducted by Uchechukwu et al. (2014) on cloud computing data centers are in general agreement with the findings obtained within the scope of this research. The study by Agrawal et al. (2012) showed that the sensitive use of computers and information technology infrastructure in an institution will result in significant energy advantages within the organization. ...
Yoğun teknoloji kullanan işletmelerin ve çalışanların çevreye verdikleri zararlar, bilişim teknolojilerindeki gelişmelerle birlikte inanılmaz boyutlara ulaşmıştır. Yeşil bilişim; işletmelerin kullandığı bilgi teknolojilerinin, bilişim kaynaklarının, tasarım ve üretim süreçlerinin, bilgisayar ve diğer teknolojik ürünlerin çevreye ve doğal kaynaklara en az zarar vermesini sağlayan bir teknoloji bilincidir. Bir diğer ifadeyle bu kavram; bilgisayar, yazıcı, monitör, depolama araçları gibi bilişim ürünlerinin çevreye en az tahribat yapacak biçimde daha etkili ve verimli olarak kullanılmasını ifade etmektedir. Dijital çağın işletmelerinin ve çalışanlarının çevreye verdikleri bu tip zararların azaltılabilmesi için, işletme içinde yeşil bilişim bilincinin yerleşmesine imkân sağlayacak yönetim stratejilerine ve bu stratejilerin örgüt politikaları çerçevesinde önceliklendirilerek sistematik bir şekilde yaygınlaştırılmasına ihtiyaç vardır. Özellikle yoğun teknoloji kullanan işletmelerin gelecek öngörülerini yeşil bilişim stratejileri doğrultusunda yapması artık bir zorunluluk haline gelmiş olup, işletmelerin bu stratejileri örgüt bünyesinde uygulaması ve yeşil kalkınmaya yönelik sürdürülebilir politikaları oluşturması gerekmektedir. Araştırmanın temel amacı; günümüz dünyasında bir zorunluluk haline gelen yeşil bilişim kavramına dikkat çekerek, işletmelerin yeşil bilişim stratejilerini nasıl geliştirebileceklerine ve hangi yöntemlerle önceliklendirebileceklerine yönelik örnek bir bakış açısı oluşturmaktır. Bu amaç doğrultusunda yeşil bilim kavramıyla ilgili araştırmalar incelenmiş ve işletmeler için en uygun yeşil bilişim stratejisini Analitik Hiyerarşi Prosesi (AHP) yöntemiyle belirlemeye yönelik bir model oluşturulmuştur. Modeldeki 4 ana kriter; yazılım, donanım, derleyiciler ve veri merkezi olarak belirlenmiştir. Ana kriterler ve her bir ana kriter içindeki alt kriterler temelinde önerilen 5 yönetim stratejisi, çalışan ve akademisyen görüşleriyle önceliklendirilerek, işletmeler için örnek oluşturabilecek bazı bulgulara ulaşılmıştır. Araştırma sonuçlarının ve ortaya koyulan önerilerin; işletmelerin yeşil bilişim stratejileri oluşturmasına, uygulamasına ve yeşil bilişim bilincine katkı sağlayabileceği düşünülmüştür.
... An important aspect of computation is the energy consumption of task execution. We focus on the dynamic energy consumption of task execution [38]. Other energy consumption on the RSU, such as energy consumption at idle mode and cooling system, is assumed constant. ...
Full-text available
Vehicular Edge Computing (VEC) systems exploit resources on both vehicles and Roadside Units (RSUs) to provide services for real-time vehicular applications that cannot be completed in the vehicles alone. Two types of decisions are critical for VEC: one is for task offloading to migrate vehicular tasks to suitable RSUs, and the other is for resource allocation at the RSUs to provide the optimal amount of computational resource to the migrated tasks under constraints on response time and energy consumption. Most of the published optimization-based methods determine the optimal solutions of the two types of decisions jointly within one optimization problem at RSUs, but the complexity of solving the optimization problem is extraordinary, because the problem is not convex and has discrete variables. Meanwhile, the nature of centralized solutions requires extra information exchange between vehicles and RSUs, which is challenged by the additional communication delay and security issues. The contribution of this paper is to decompose the joint optimization problem into two decoupled subproblems: task offloading and resource allocation. Both subproblems are reformulated for efficient solutions. The resource allocation problem is simplified by dual decomposition and can be solved at vehicles in a decentralized way. The task offloading problem is transformed from a discrete problem to a continuous convex one by a probability-based solution. Our new method efficiently achieves a near-optimal solution through decentralized optimizations, and the error bound between the solution and the true optimum is analyzed. Simulation results demonstrate the advantage of the proposed approach.
... In literature, it observed that idle servers consume 70 percent of maximum power consumption. An average of 40 percent of total power consumption observed from cooling equipment and usage of CPU, memory, storage, and network interfaces contributes to total consumption [24]. In [23] authors proposed an energy model that also considers these parameters and defined the Energy model as: ...
Full-text available
Resource optimization is becoming a prime factor in the progress of Internet-based technology, Cloud Computing. A resource management model is highly required in cloud data center paradigms to utilize available resources effectively. Bin-Packing problem is an applicable combinatorial optimization for Virtual Machine (VM) to Physical Machine (PM) allocation to minimize the required PMs. In this paper, we have proposed an efficient resource allocation and management algorithm in two phases. During the first phase, a Load Balanced Multi-Dimensional Bin-Packing (LBMBP) heuristic for Virtual Machine (VM) to Physical Machine (PMs or host) allocation is introduced, considering multidimensional resources: CPU, RAM, and Network Bandwidth. In the Second Phase, to perform VM migration, a mechanism to detect overloaded and underloaded hosts based on outliers has been described. The proposed work illustrated the simulation results using CloudSim Plus Simulator and observed a reduction in the number of active PMs. Energy consumption and the number of migrations with improved resource utilization.
... The delay occurred due to centralized processing of IoT tasks may leads to serious problem if a patient is waiting for some result. The other limitations of cloud computing are low bandwidth, energy consumption, etc. [19]. In recent years, smart healthcare becomes an emerging IoT application. ...
Full-text available
The advancement of internet technology along with the high adoption rate of various IoT devices with various emergent IoT applications has put stringent Quality of Service (QoS) and Quality of Experience (QoE) requirements on the service provider. The existing IoT-Cloud model cannot cater the QoS and QoE requirements of these IoT applications effectively and efficiently due to low bandwidth and high propagation delay between the IoT device layer to the cloud data center. To fulfill the aforementioned demand of these IoT applications Fog Computing has been used as an extension of cloud computing to provide multiple IoT services to the end user at the close vicinity of the IoT devices. Fog computing paradigm is used to support various types of IoT services through hardware and software virtualization in distributed and heterogeneous manner to execute different IoT applications at the edge of the network. The request generated by the respective IoT devices are mostly random in nature, and the objective of these applications varies dynamically hence finding the optimal fog resources for the processing and execution is a challenge. In order to find the optimal mapping between the IoT applications and Fog resource different resource allocation policy has been studied like resource scheduling, task scheduling, task placement etc. In this article we have formulated optimization model for multiple healthcare IoT application placement problem in Fog computing architecture.
Full-text available
This paper presents a means to quantify air flow performance of both traditional raisedfloor designs and non-raised floor air-conditioning designs for data centers. Metrics are developed that will readily perm it owners, engineers and operators to measure and quantify the effectiveness of their data center cooling systems or changes that they make to their cooling systems to increase air cooling efficiency. The metrics incorporate and integrate together the major factors that decrease the effectiveness of computer room air cooling. These factors, which are covered in the paper, include; negative pressure flow rate (air induced into the floor void), bypass flow rate (from floor void directly back to the air-conditioners without cooling servers), recirculation flow rate (from server outlet, back into server inlet) and the balance of CRAC and server design flow rates. Examples of the application of these metrics are also presented.
Full-text available
In this paper, we propose two energy models based on a statistical analysis of a server's operational behavior in order to minimize the energy consumption in data centers at cloud computing providers. Based on these models, the Energy Savings Engine (ESE) in the cloud provider decides either to migrate the virtual machines (VMs) from a lightly-loaded server and then turn it off or put it in a sleep mode, or to keep the current server running and ready to receive any new load requests. The main difference between the two models is the energy and time required to put the server in operational mode from a sleep mode or from an off state. Therefore, the decision is a tradeoff between the energy savings and the required performance according to the SLA between the client and the cloud provider. We show results based on actual power measurements taken at the server's AC input, to determine the energy consumed in the idle state, the sleep state, the off state and in the case of switching between any two of these states. In addition, we measured the power consumed by the source and the destination servers during the migration of a VM.
The dimensionless capture index (CI) is proposed as a cooling performance metric based solely on the airflow patterns associated with the supply of cool air to, or the removal of hot air from, a rack. The capture index is typically a rack-by-rack metric and has values between zero and 100%; higher values generally imply good cooling performance and scalability of a cooling architecture. In many applications, the capture index provides additional information relative to rack-inlet temperatures and other cooling indices. In some applications, capture index may be computed in lieu of other metrics. Two variants of capture index are defined: one for cold- and another for hot-aisle analyses. A technique for computing CIs numerically, based on tracking airflow with passive concentrations, is provided along with examples based on CFD simulations. A cluster-wide cooling performance metric based on CI, the total escaped power, is also proposed. Finally, an example shows how CI and total escaped power may be used alongside other metrics to determine an optimized physical arrangement of a given inventory of racks and coolers bounding a hot aisle.
Conference Paper
Cloud computing delivers computing as a utility to users worldwide. A consequence of this model is that cloud data centres have high deployment and operational costs, as well as significant carbon footprints for the environment. We need to develop Green Cloud Computing (GCC) solutions that reduce these deployment and operational costs and thus save energy and reduce adverse environmental impacts. In order to achieve this objective, a thorough understanding of the energy consumption patterns in complex Cloud environments is needed. We present a new energy consumption model and associated analysis tool for Cloud computing environments. We measure energy consumption in Cloud environments based on different runtime tasks. Empirical analysis of the correlation of energy consumption and Cloud data and computational tasks, as well as system performance, will be investigated based on our energy consumption model and analysis tool. Our research results can be integrated into Cloud systems to monitor energy consumption and support static or dynamic system-level optimisation.
The greatest environmental challenge today is global warming, which is caused by carbon emissions. A report by the Energy Information Administration says that about 98 percent of CO2 emissions (or 87 percent of all CO2- equivalent emissions from all greenhouse gases) can be directly attributed to energy consumption. The major challenge of many organizations today is a desire to operate in a “green” manner, publishing principles for environmental practices and sustainability on their corporate Web. In addition, many companies are now paying some kind of carbon tax for the resources they consume and the environmental impact of the products and services they produce, so a reduction in energy consumed can have a real financial payback. In this paper, we focus on reduction in energy consumption over the full equipment life cycle as the prime motivator for “green” application design; with energy reduction as the best measure of “green-ness”. Green IT refers to the study and practice of using computing resources in an efficient, effective and economic way. The various approaches of the green IT are Virtualization, Power management, Material recycling and Telecommuting. In which Virtualization platforms can run across hundreds of interconnected physical computers and storage devices, to create an entire virtual infrastructure. Cloud computing is the concept of equal sharing of resources among nodes. Cloud virtualization refers the process of running two or more logical computer systems on one set of hardware with equal sharing of resources. The challenge of the existing system is reduction of efficiency due server virtualization. In this paper, we propose a new system that connects more number of nodes with minimum number of servers. The sole motivation of this paper is reducing energy consumption using cloud virtualization clique star cover number technique.