Shuja et al.

Journal of Internet Services and Applications (2017) 8:9

DOI 10.1186/s13174-017-0060-5
Journal of Internet Services
and Applications

RESEARCH Open Access

Greening emerging IT technologies:

techniques and practices
Junaid Shuja1*† , Raja Wasim Ahmad1† , Abdullah Gani2 , Abdelmuttlib Ibrahim Abdalla Ahmed2 ,
Aisha Siddiqa2 , Kashif Nisar3 , Samee U. Khan4 and Albert Y. Zomaya5

Abstract
The tremendous increase in global industrial activity has resulted in high utilization of natural energy resources and
increase in global warming over the last few decades. Meanwhile, computing has become a popular utility of modern
human lifestyle. With the increased popularity of computing and IT services, the corresponding energy consumption
of the IT industry has also increased rapidly. The computing community realizes the importance of green measures
and provides technological solutions that lead to its energy-aware operations along with facilitating the same in other
IT enabled industries. Green and sustainable computing practices review the environmental impact of the computing
industry to encourage the adoption of practices and technologies for efficient operations. “Green Computing”
paradigm advocates the energy-proportional and efficient usage of computing resources in all emerging
technologies, such as Big Data and Internet of Things (IoT). This article presents a review of green computing
techniques amidst the emerging IT technologies that are evident in our society. The best practices for green
computing and the trade-off between green and high-performance policies is debated. Further, we discuss the
imminent challenges facing the efficient green operations of emerging IT technologies.
Keywords: Green computing, IT services, Big data, IoT, Cloud computing, Mobile computing, Software defined
networks

1 Introduction IT technologies, existing practices, and algorithms need

Global industrial growth has had two demanding to be redefined for energy efficient, energy-proportional,
effects on the human environment. First, natural energy and sustainable operations. Additionally, IT technologies
resources are being consumed at a rapid pace. Efficient have a responsibility to limit the energy consumption and
operations and alternate energy resources are sought to carbon footprint of other industries and organizations
reduce the current rate of depletion of natural energy while facilitating green environmental practices in their
resources. Second, global industrial growth has resulted daily operations [1].
in increased carbon emissions. The carbon emissions, Many modern aspects of our society are based on the
known as Greenhouse Gases (GHG) in general, lead to global success of the IT industry. The problem of energy
higher disease rates, global warming, and depletion of and sustainability is often associated with manufactur-
the Ozone layer. Information Technology (IT) is both an ing, aviation, and petroleum industries. However, the IT
emerging global industry and a support technology for sector is also accountable for high energy consumption
many businesses. We seek information at increasing rates and carbon emissions. The IT sector is currently respon-
and in multiple forms to ease our lifestyle. The IT industry, sible for 2.4-3% of global electricity consumption with
or computing in general, contributes to both high energy aforecasted 20% increase annually. Similarly, the IT sec-
consumption and carbon emissions. Therefore, emerging tor accounts for 2-2.5% of worldwide carbon emissions
equivalent to 0.86 metric gigatonnes of CO2 [2]. The
*Correspondence: [email protected] increasing energy and carbon impact of computing call for
† Equal contributors
1
COMSATS Institute of Information Technology, University Road, Abbottabad, energy-proportional and “Green” computing systems.
Pakistan Green Computing is a computing paradigm where: (a)
Full list of author information is available at the end of the article IT resource efficiencies are maximized, (b) resources

© The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and
reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons license, and indicate if changes were made.
Shuja et al. Journal of Internet Services and Applications (2017) 8:9 Page 2 of 11

(in particular, energy) are re-used whenever possible, cloud computing resources to process the large amount of
(c) sustainable products and manufacturing practices are data collected through edge-based devices (smartphones,
adopted, and (d) green initiatives in other industries are cameras, and sensors) [7]. International Data Corpora-
supported through monitoring and management tools [3]. tion (IDC) annual reports also point to the increasing
Resource efficiency has dual context for performance and revenue and usage of cloud computing, mobile comput-
energy. Multiple resources and their alternative backups ing, IoT, big data analytics and software-based networks
are utilized for efficient high-performance computing. In technologies [8, 9].
contrast, energy efficiency practices involve reduction of The rest of the article is structured as follows.
resources and energy proportional computing. Energy re- Sections 2, 3, 4, 5, and 6 discuss the greening of cloud
use in computing systems derives from the cyber-physical computing, mobile computing, IoT, Big data analytics,
interactions of IT resources and cooling of large-scale and software-based networks respectively. In Section 7,
IT centers [1]. Similarly, sustainability in computing is research issues and challenges to the green computing
achieved by utilization of renewable energy resources paradigm and emerging IT technologies are listed. We
while limiting the carbon footprint of IT operations. Man- provide concluding remarks in Section 8.
ufacturing practices that increase re-use of off-the-shelf
computing components and limit the e-waste also con- 2 Green cloud computing
tribute to sustainability efforts. Other than self-conscious Cloud computing has established itself as an enabling
approaches to green computing, the IT technologies are technology for multiple IT services. The increase in
utilized as a platform to promote the greening and sus- the number of cloud-based IT services and applications
tainability efforts of other industries through environmen- demands establishment of data centers that house thou-
tal monitoring and social awareness [4]. sands of web servers, storage, and network devices. Cloud
In recent years, multiple IT technologies have inte- data centers (CDC) provide a range of services from high-
grated into people’s lifestyles seamlessly while facilitating performance computing to large-scale data analytics to
day-to-day tasks, such as social communications, health- end users. The massive scale of cloud data centers that
care monitoring, and environmental management [5]. We are setup at multiple geographical locations to facilitate
present the “Green Computing” paradigm in this article distributed users means that they contribute 25% to the
from the perspective of emerging IT technologies and total IT electricity share [10]. Moreover, IT services are
their green initiatives. We select (a) cloud computing, (b) shifting from single server operations to rack-mounted
mobile computing, (c) Internet of Things (IoT), (d) big blade servers. The rack-mounted server designs result in
data analytics, and (d) software-based networks as emerg- higher electronic densities, higher energy consumption,
ing IT technologies. This article is different from previous and heat dissipation [11]. As a result, both direct energy
efforts on summarizing the green technologies in segre- and indirect cooling energy demands rise in cloud data
gated IT technologies. The contributions of this article centers. The techniques to “green” cloud data center oper-
are (a) we provide a survey on green algorithms, circuits, ations can be broadly classified into three categories: (a)
architectures, and practices in emerging IT technologies resource management with virtualization, (b) sustainabil-
of this decade, (b) we highlight the key requirements and ity with renewable energy and waste heat utilization, (c)
practices for the greening of emerging IT technologies, and resource scheduling with state-of-the-art evolution-
and (c) we emphasize future research trends in the field of ary algorithms [12].
green computing. Cloud data center resources are managed by a virtual-
Multiple computing technologies have emerged over ization layer that resides over the physical resources. The
the last decade as enablers of scientific, industrial, and virtualization layer abstracts the hardware layer interfaces
social businesses. To select the emerging IT technolo- to provide a higher level interface for users and appli-
gies for a discussion on green computing, we set met- cations. The virtualization layer helps in management
rics on popularity, social integration, and future appli- and consolidation of cloud data center resources through
cation in smart environments [6]. We chose cloud com- multiple backup techniques, such as resource migration
puting, mobile computing, IoT, big data analytics, and and snapshot [13]. The primary objective of virtualization
software-based networks as the emerging IT technologies in cloud data centers is to provide scalable and fault-
based on the aforementioned metrics. The integration and tolerant operations. Increasingly, virtualization is being
high correlation of these technologies create opportuni- used for resource consolidation and energy efficiency. A
ties that assist various organizations in performing their virtual resource residing on a 40% utilized server can be
duties efficiently. For instance, due to high integration of migrated to another 40% utilized server while the for-
the technologies, law enforcement agencies were able to mer is operated in low-power idle mode [10]. The virtual
detect the bomber behind the Boston marathon bomb- machine (VM) migration is exploited in both inter and
ing in four days. During the investigation, the FBI utilized intra-data center configurations while providing energy
Shuja et al. Journal of Internet Services and Applications (2017) 8:9 Page 3 of 11

efficient operations. However, the network cost result- modeled as graph and tree-based structures. The resultant
ing from the VM migration needs to be addressed for optimization model of the cloud services generally focuses
joint network and server resource optimization [14]. The on the task makespan and cost minimization while deter-
intra-data center VM migration network cost is reduced mining which task is allocated to which resource [21].
by placing related and “talkative” VMs in optimal server Recently, with a shift in focus on the energy consumption,
proximity so that their communications are limited to a the energy costs have been included in the optimization
part of the network [15]. Similarly, the inter-data center models. However, the task makespan and energy mini-
VM migration cost is reduced by data deduplication and mization requirements often conflict. The task makespan
compression techniques over long-haul networks [16]. minimization requires exploitation of multiple resources
The green computing initiative also embodies sustain- that leads to higher energy costs. The multi-objective
ability in operations. Cloud data centers operating on modeling of the task allocation problem in cloud data cen-
renewable energy resources lead to zero GHG emis- ters with thousands of resources and applications leads
sions. Renewable energy from sources such as the sun to great complexity in solution finding [22]. Evolution-
and the wind can be generated from on-site installations ary algorithms are employed to swiftly find near-optimal
or purchased from off-site corporations [4]. The main solutions for the multi-objective energy efficient resource
drawbacks of renewable energy based cloud data cen- scheduling problems in cloud data centers [3]. Interested
ter operations are the associated cost and unpredictable readers can refer to comprehensive surveys on the green-
supply of the renewable resource. It is estimated that ing of cloud computing [1, 21]. Figure 1 illustrates the
with the advances in storage capacities, the cost/Watt of options for greening cloud computing systems.
renewable energy will halve in the next decade [17]. More-
over, to address the unpredictability of renewable energy 3 Green mobile computing
resources, techniques such as dynamic power-workload Smartphones of recent generations are equipped with
balancing and server power capping are exploited [1]. high storage capacity and the computational power to per-
The integration of renewable energy resources to cloud form resource-intensive tasks. The preference of smart-
data center power designs requires utilization of hybrid phone users has lessened the dependency on desk-
power supplies and Autonomic Transfer Switches (ATS). top servers to perform computing tasks. As a result,
The ATS shift power between grid and renewable energy the resource requirements of the smartphone applica-
resources to match the dynamic data center workload with tions have also increased [23]. Emerging media-rich
the power generation [18]. smartphone applications frequently trigger sensors, such
The re-use of resources is a major goal of the green as GPS, accelerometer, and wireless radios to provide
computing paradigm. Modern modular data centers with context-aware services. As a result, the computation, com-
blade servers of higher electronic densities are leading to munication, and energy cost of smartphones significantly
increased cooling requirements. It is estimated that 40% increase. To handle the energy-performance trade-off,
of data center electricity is used in cooling the servers energy-efficient system designs are necessary to meet
while keeping their temperatures in operational range the requirements of modern smartphone devices. More-
[19]. The waste heat generated by data centers can be over, energy estimation techniques help to propose the
utilized or re-used in various waste heat recovery sce- energy-efficient design of smartphone applications and
narios. Firstly, cloud data centers provide ample oppor-
tunity for waste heat re-use in the cooling process. The
heat recovered from servers is captured in the vapor-
absorption based cooling systems where reversible heat
pumps transfer thermal energy to cooler space. Secondly,
in cooler places, data centers can be co-located with res-
idential buildings for district heating [20]. Thirdly, mod-
ular data center designs can be migrated to cooler areas
to reduce the cooling requirements while directly utiliz-
ing ambient air in the cooling process [1]. The major
concern with energy re-use in CDCs is the low quality
of heat generated that is applicable to few heat recovery
processes [11].
The basic objective of cloud data centers is to pro-
vide IT services at an optimal pay-as-you-go model. In
most software-based green cloud computing solutions, Fig. 1 Green Cloud Computing Techniques
the network, processor, storage devices, and user tasks are
Shuja et al. Journal of Internet Services and Applications (2017) 8:9 Page 4 of 11

system components. Energy estimation helps to identify into a sleep state that consequently depletes mobile bat-
the rogue applications within a smartphone [24]. tery charge. A mobile application, with no sleep bug,
Effective management of the hardware components acquires a lock on a mobile component and does not
of a smartphone device significantly improves the total release it for a long period of time. The ADEL framework
energy budget. The architectural design of hardware mod- reported energy bugs of Wi-Fi components by tracking
ules within the smartphone is based on Complementary the packet transmission rate within the mobile application
Metal–Oxide–Semiconductor (CMOS). The total power using dynamic taint-tracking analysis. Handling energy
consumption of CMOS based circuits (e.g. CPU, static bugs puts some extra burden on programmers to explic-
RAM, and GPU) consists of static and dynamic power. itly manipulate power control APIs for energy-efficient
The static power of a circuit varies from device to device operations of mobile applications [24].
depending on the insulation capabilities of transistors and Smartphone energy estimation provides the basis for
represents the power consumption when the transistor green computing within smartphones. It provides feed-
is not in the switching state [25]. Dynamic power rep- back to the application developers to consider energy as
resents the power consumption when a device changes a metric in addition to maintainability, complexity, and
logic state from on to off or vice versa. Power gat- understandability. Smartphone application energy esti-
ing embeds a high voltage threshold transistor between mation schemes are broadly classified into components
actual ground and circuit ground of a device to switch power model and code analysis based estimation cat-
off the transistor during its sleep hours to reduce leak- egories. Component power model based methods use
age power. For the CPU module, dynamic frequency State of Charge (SOC) estimation methods to forecast
scaling (DFS) enables dynamic adjustment of power con- the energy consumption of an application [24, 30]. Alter-
sumption for greening the smartphones at the cost of natively, the code analysis based method considers base
throughput [26]. The tail power represents the state of a cost energy of instructions within the source code of an
smartphone component that remains in high power state application to estimate energy consumption [31]. SOC
although it has already finished its required task. The tail estimation methods include coulomb counting and volt-
power state of smartphone components such as Wi-Fi, age based methods. Coulomb counting estimates SOC
3G, GPS, and SD-CARD, depletes battery charge quickly. by communicating to the smartphone’s built-in sensors
Software tools, such as E-prof, empower smartphones to to find the accumulative current drop rate over time.
measure/estimate the device energy consumption at the However, coulomb counting produces inaccurate esti-
component level. However, software-based solutions sig- mation results due to internal factors such as battery
nificantly impact the device’s energy consumption due to aging, the temperature within the smartphone, and charg-
their profiling activities [27]. ing/discharging rate. Alternatively, voltage-based SOC
Software based green computing solutions such as estimation employs fuel gauge sensors. Fuel gauge sen-
mobile cloud computing based computational offload- sors are inaccurate owing to low charge update rate. Base
ing, energy bug handling, and energy efficient application cost energy methods assign base cost to the operations
development significantly reduce the energy budget of the within an instruction to estimate energy consumption of
smartphone. Mobile cloud computing empowers smart- an application based on static code analysis. The esti-
phone devices to augment device lifetime by carefully mation method helps either to improve the hardware
offloading energy critical tasks to remote cloud servers. components of smartphones or software for green com-
Computational offloading decisions consider total execu- puting [31]. Figure 2 highlights the hardware and software
tion time, resource consumption, energy requirements, options for green mobile computing.
and privacy issues of an application before migrating Inefficient code design within a smartphone applica-
a task to resource-rich cloud servers [28]. Energy bugs tion has a high impact on the total energy consumption.
within a smartphone lead to abnormal power consump- Within an application, resource optimal placement of
tion behavior of mobile applications. Energy bugs are classes and functions reduces the power consumption.
difficult to track, and mainly occur due to (a) faulty bat- For instance, minimizing the memory distance between
teries, (b) damaged mobile battery chargers, (c) infected two functions that frequently communicate reduces the
memory cards, and (d) damaged SIM cards. Alternatively, energy consumption of target application [32]. Also,
within an OS, changing OS configuration impacts the educating developers with energy efficient application
mobile battery power consumption rate. For instance, set- development techniques including loop unrolling, branch
ting SetCPU function incorrectly for kernel overclocking optimization, dead store elimination, value numbering,
results in high battery power consumption [29]. Simi- code inlining, constant propagation, code motion, inter-
larly, infected mobile applications and frameworks also procedural analysis, and instruction scheduling, greens
drain mobile application abnormally. For instance, a “no smartphone operations [33]. Extensive studies on green
sleep” bug hinders a smartphone component from going mobile computing are listed for detailed analysis [24, 34].
Shuja et al. Journal of Internet Services and Applications (2017) 8:9 Page 5 of 11

tag reader triggers information flow through transmitting

a query signal. Consequently, the responses come from
the nearby RFID tags. Mostly, RFID system transmission
ranges are not more than a few meters. Moreover, the
transmission frequencies start at 124-135 kHz up to
ultra-high at 860-960 MHz. Currently, RFID tags can be
found in two types: active and passive tags. The active
tag uses onboard power batteries to do its functions. The
passive tags depend on harvesting energy from the signal
of the readers following the principle of induction [36].
To obtain green RFID, two factors should be considered.
Firstly, RFID tag sizes should be reduced since tags them-
Fig. 2 Green Mobile Computing. The figure depicts techniques for selves are difficult to recycle. Consequently, the amount
green mobile computing of non-degradable material should be reduced in tag man-
ufacturing (e.g. printable RFID tags, paper-based RFID
tags, and biodegradable RFID tags). Secondly, using com-
4 Green internet of things munication algorithms and protocols that support energy
IoT is another emerging technology that facilitates data efficiency can lead to Green IoT. Green communica-
communication among multiple electronic devices with- tion protocols provide energy efficiency through dynamic
out human and computer intervention. Green IoT is a set adjustment of the level of transmission power, optimiza-
of procedures adopted by the IoT in the form of hard- tion of tag estimation, and avoiding of tag collision and
ware or software efficiency techniques. Green IoT aims overhearing [37].
to achieve energy efficiency through the reduction of the Green Wireless Sensor Network (GWSN) A Wire-
greenhouse effect in the current services and applications. less Sensor Network (WSN) comprise of numerous sen-
Moreover, to reduce the impact on the environment, sor nodes that have resource-constraints, such as limited
Green IoT focuses on the issues of green productions, computing capability, storage capacity, and power. Com-
green redesign, and green recycling/disposal [35]. Table 1 monly, the sensor nodes are connected to a powerful base
highlights enabling technologies and greening strategies station called sink. Usually, sensor nodes are equipped
for IoT. with multiple on-board sensors to read the surroundings
Real deployment of IoT is performed through the collab- circumstances, such as humidity, temperature, accelera-
oration of enabling technologies, communication strate- tion, etc. Commercial WSN solutions are based on the
gies, and protocols. This section mainly focuses on the IEEE 802.15.4 standard [38]. Techniques such as (a) sleep
most crucial communication strategies and technologies mode activation during sensor idle time, (b) wireless
that lead towards green IoT. charging mechanisms that harvest environmental mech-
Green Radio-Frequency IDentification (GRFID) anisms, (c) radio optimization, and (d) energy efficient
Radio Frequency Identification (RFID) is one of the routing and data collection are utilized for GWSN [39].
promising IoT enablers. A RFID system comprises of WSN aggregate sensed data into a sink from cluster
RFID tags and tags readers. RFID tags are in the form heads through event-detection and continuous monitor-
of microchips attached to the radio that works as a ing. Cluster heads receive and send aggregated data con-
transceiver. Every RFID tag has a unique ID and can store tinuously, which leads to faster energy depletion around
context data regarding the entities to which they are the sink [40]. There are two strategies for optimizing
attached. Generally, in the elementary process, the RFID energy usage in WSN, namely, periodic reporting instead

Table 1 IoT Enabling technologies and their greening strategies

IoT Enabler Type Communication Data transfer Power Life-time Greening strategies
Paradigm source
Active tags Two ways Low Battery ≤5
RFID Energy-efficient algorithms and protocols
Passive tags One way Very low Harvested ∞

Smart object Central Low Battery ≤5

Sensing network Sleep wake-up, data reduction mechanisms
Mobile sensing P2P High Battery ≤2

Cloud Client/Server Very low Grid ≤ 10yrs Turn off unrequired facilities, Minimize data
Internet technologies
Future internet Distributed Very low Grid ∞ path length
Shuja et al. Journal of Internet Services and Applications (2017) 8:9 Page 6 of 11

of continuous monitoring and timestamp-less synchro- literature frequently. Cloud computing is revealed as a
nization. In a periodic reporting strategy, time periods big data analytics technology which offers resource out-
of data reporting are set by the sensor owner to avoid sourcing in order to avoid physical occupation and thus
energy spikes that are raised in event-driven reporting. multiple users with varying analytics requirements can
In timestamp-less synchronization, the broadcast con- utilize remotely accessible resources. The advancement in
trol messages to sensors for synchronization are not cloud computing for big data analytics is expected to lead
put to practice. The participating sensor pairs performs to low dependency on the usage of personal computers in
REQUEST/RESPONSE until the achievement of synchro- the new era of computing. Along with resource preserva-
nization process [41]. tion, cloud computing also offers lower energy consump-
Green M2M communication(GM2MC) Machine-to- tion for executing high computational procedures on big
machine (M2M) communication is one of the popular data [45]. Cloud computing has great importance as being
paradigms in IoT. There are two communication domains a highly available platform for big data analytics which
in IoT: M2M and networks. In an M2M domain, multi- allows minimization in resource utilization and energy
ple nodes are deployed to intelligently monitor and gather consumption [46].
data. In the network domain, wireless/wired networks There is a visible advancement in today’s technology
carry the gathered data to the desired base station (BS). towards green big data analytics. For instance, Green-
The BS supports different M2M applications through the Plum [47] and GreenHadoop [48] are proposed in big
network. The challenge is that the massive nodes involved data analytics for green computing. GreenPlum is an open
in M2M interactions consume a lot of energy. The tech- source data warehouse, licensed under Apache Inc., which
niques that can be utilized to improve energy-efficiency offers fast analytics on petabyte-scale data with efficient
of M2M communications are: (a) intelligently adjusting query processing via parallel processing and optimization.
the transmission power to the necessary level, (b) devel- Cost-based query optimization introduced by GreenPlum
oping energy-efficient routing protocols, (c) scheduling ensures high analytics on large volume data sets with
the activity in the machine domain, and (d) using energy- usage efficiency. GreenHadoop, on the other hand, brings
harvesting techniques [42]. Zhu et al. [39] provide exhaus- the idea of renewable energy sources in order to bal-
tive reading on Green IoT technologies. ance the supply and demand of energy sources associated
with big data analytics. The GreenHadoop framework
5 Green big data analytics uses a photovoltaic solar array and electrical grid energy
Big data introduces the era of data with new challenges resources. The GreenHadoop framework for green analyt-
such as petabyte scale structured and unstructured data ics achieves maximized energy consumption by estimat-
sets which are growing at an exponential rate and have ing available solar energy and scheduling MapReduce jobs
heterogeneous formats. Fast data retrieval and accuracy accordingly. GreenPlum provides support to both batch
of search from a pool of big data are the main chal- and interactive modes of processing. However, Green-
lenges to maximize value for decision making in big data Hadoop achieves real-time energy estimates based on
analytics [43]. Traditional data management systems lack prior data center workload.
the capability to handle big data storage and analytics Figure 3 shows a green big data analytics process where
requirements and thus NoSQL technology is contribut- storage and processing resources reside on clouds and can
ing to provide suitable solutions for timely data retrieval be requested on demand. Cloud computing technology
and efficient data processing. The process of greening is provides the basis for green big data analytics as the opti-
crucial for big data as analytics on tremendous size of mum resource utilization with reduced energy consump-
data sets requires high computing power, scalable and effi- tion. Currently, major big data sources and consumers are
cient storage space, high availability of main memory, and social networks, healthcare, industries, commerce, and
fast communication media on always-on local physical business enterprises. Data from these sources and con-
or enterprise cloud servers [44]. Consequently, green big sumers is extensively scalable and brings critical analytics
data analytics requires efficiency in resource utilization, requirements for timely decision making. This big data
energy consumption, and infrastructure scalability. storage and processing load are efficiently handled by
Big data analytics procedures may contribute to pre- data centers and processors residing on the cloud which
serving the usage of processing and storage resources, ensures green analytics. According to a study [49], it is
scalability of systems, and improved productivity. Big data estimated that cloud computing will be able to achieve
analytics requirements such as high availability, reliabil- 38% reduction in energy usage by 2020. The concept of
ity, and consistency are significant in the development of recycling is stated in [50] which suggests that renewable
technological infrastructures. However, energy preserv- energy technology will be a preferable choice of invest-
ing and resource optimization are the green computing ment in finding energy resources by 2040. Renewable
aspects of analytics which have not being reported in the energy technology is emerging with reduced adaptation
Shuja et al. Journal of Internet Services and Applications (2017) 8:9 Page 7 of 11

Fig. 3 Green Big Data Analytics. The figure depicts techniques for green big data analytics

cost, efficient green housing, and increased renewability hand-off and (b) Adaptive Link Rate (ALR) techniques
demands which aim to achieve reduced carbon discharge, that scale link rate and utilize sleep states for energy-
lower and stable energy costs, and access to reliable energy proportional computing [55].
sources. 64% of the IT industry are meeting their targets of The software and virtualization techniques have led to
energy saving by using renewable energy technology [51]. current advancements in the energy efficiency of net-
Green big data analytics is significant in optimizing working technologies. Software Defined Networks (SDN)
energy consumption and re-usability of available sources separate the data and control plane of network routers
to meet extensive analytics requirements of big data. with the help of a central controller. SDN do not have a
Green computing is analogous to green chemistry and direct impact on the energy consumption of a network.
allows usage minimization for enormous computing and However, the pervasive programmable interface of SDN
storage resources required by big data. Green computing supports energy efficient network operations indirectly
aligns the big data analytics technologies with the concept through resource consolidation [56]. A minimum energy
of sustainability i.e. reduction, reusability, and recycling. efficient subset of network resources can be calculated
Researchers [51] suggest that the technology industry through a resource optimization technique and imple-
seems more concerned about analytics efficiency than mented through SDN as demonstrated in [57]. Hence,
environmental sustainability and computational complex- server and network resource management techniques can
ity. However, implementation of green analytics on big be utilized in parallel with the virtualization and SDN
data surely results in reduced memory usage and compu- enabling technologies. SDN can help implement green
tational cost. Interested readers can refer to an extensive computing policies at the network level based on their
future perspective on green big data analytics [52, 53]. programmable control plane. Similarly, security policies
can be implemented with the help of SDN while elim-
6 Green networking inating the need for stand-alone security devices. Con-
Networks are the basic component and enabler of the sequently, SDN-enabled network devices can also imple-
innovations that have occurred in human society in the ment security functions, lowering the total operational
past few decades. As more industries and business have costs and energy bill [58].
integrated IT technologies and services, the networks Network Function Virtualization (NFV) is another
have grown into complex structures connecting billions technological shift in telecommunication systems. NFV
of devices worldwide. As a result, network devices con- decouples network forwarding and routing functions
sume a large amount of energy constituting approximately from underlying physical systems through virtualization
10% of the aggregate IT energy consumption [54]. The [59]. Network functions, such as a firewall, can be imple-
basic techniques applied for energy efficient networks are: mented in software (virtual network function) and imple-
(a) energy efficient protocols for routing, medium access, mented on any of the industry standard physical servers.
Shuja et al. Journal of Internet Services and Applications (2017) 8:9 Page 8 of 11

Similarly, network devices can offer virtual computa- Green cloud computing: Green cloud computing
tion services. As a result of virtualization, network and demands divergence from conventional computing tech-
compute devices offer agile computing and forwarding niques, hence, increased operational and infrastructural
functions reducing the capital and operational costs of costs. For example, renewable energy has a higher cost
all IT services, especially cloud computing. The decou- than conventional grid energy. Similarly, waste heat uti-
pling of network functions from physical devices results in lization measures in data centers also demand costly ther-
flexible and dynamic resource scheduling, hence, energy mal heat exchange materials. Incorporating green mea-
efficiency [60]. Five out of six case studies show that the sures with cost-efficient business operations is a challeng-
NFV based networks provide energy savings compared ing task in cloud data centers. The efficiency of renewable
to baseline networks. Similarly, higher performance and energy generation and storage mediums needs to be rigor-
energy efficiency were observed as compared to com- ously increased in order to provide comparable business
modity servers while experimenting with a virtualized incentives. The cost of VM migrations for resource con-
Deep Packet Inspection (DPI) application [59]. However, solidation over long-haul networks is also a highly debated
a balance between network function performance and research issue [14, 65]. Moreover, government policies
energy efficiency achieved through virtualization has to be need to be devised that provide incentives to green cloud
resolved. computing business providers and users.
Both SDN and NFV technologies are in early stages of Green mobile computing: Mobile application energy
deployment. Therefore, research on the development of optimization demands precise estimation accuracy for
green computing architectures based on SDN and NFV efficient battery resource usage. Empowering applica-
technologies has significant future prospects in terms of tion developers with a fine granular energy estimation
integration with other IT technologies. Interested read- tool to estimate the energy behavior of an application
ers can refer to [61] for a detailed survey on green at earlier development stages augments device battery
networks. lifetime. Existing energy estimation tools such as power
tutor, trepn profile, and Nokia energy profiler, run the
application on the smartphone to record power states of
7 Practices, research challenges, and issues power models for smartphone components to estimate
In this section, we debate the practices, research issues, energy consumption. However, because of low accuracy
and challenges to the green initiatives in emerging IT of fuel gauge sensors within smartphone batteries, the
technologies in particular and computing in general. estimate accuracy is limited. Also, the energy estimation
Green computing practices emphasize the implemen- time and overhead is high. To challenge the aforemen-
tation of green technologies at industrial and organi- tioned issues, there is a need to develop an estimation
zational level. The cost of per unit energy will rise tool that should offer high estimation accuracy and lim-
significantly owing to a considerable decrease in global ited estimation overhead. One possible solution to this
energy resources. As a result, it has become necessary problem is to estimate energy consumption based on
for both public and government sectors to propose and operational cost (energy and execution time) of different
practice state-of-the-art strategies and plans for green functions within the software. However, due to the non-
computing [62]. State-of-the-art green computing prac- deterministic nature of smartphone applications, estima-
tices consider implementation of energy friendly IT equip- tion accuracy is significantly affected. Moreover, software
ment, lightweight resource consumption protocols, and operational cost based estimation also requires accurate
disposal of electronic waste [63]. Green computing prac- estimation of code storage location. The weighted prob-
tices emphasize turning off IT resources when not utilized abilistic approach is a possible solution to resolve these
for an extended period of time. Green computing prac- issues [24].
tices also schedule IT resources in low system power and Green big data: Estimation and calculation of energy
idle states. The standby execution mode is applied for consumption for big data analytics is challenging. High
saving power if the execution power state is lower than and rapid analytic demands of big data are only satis-
a threshold [64]. The management of aging IT resources fied when an efficient estimation is available. Similarly, for
is another important issue in green computing. Older GreenHadoop, it is challenging to estimate the energy and
hardware devices have increased power consumption and time requirements for a job based upon which schedul-
require resource replacements and disposals. Hence, the ing decisions are made. Estimation is also significant in
practice of recycling needs to be applied to aging IT renewable energy technology and thus, requires extensive
resources. Similarly, practices limiting the utilization of work from academia and industry. Continuously increas-
paper prints should be applied at organizational level [4]. ing big data volume requires scalable increment in avail-
The research challenges to emerging IT technologies are able analytic resources and cost. However, the concept
listed in the paragraphs below. of green computing suggests sustainability of energy and
Shuja et al. Journal of Internet Services and Applications (2017) 8:9 Page 9 of 11

processing resources. Consequently, big data analytics of green computing practices by IT industries and IT
technology with minimized impact on the environment is empowered businesses.
highly desirable [43].
Green IoT: To preserve Green IoT some challenges 8 Conclusion
arise and need to be addressed such as Green IoT Archi- In this article, a review of the Green Computing paradigm
tectures, Green Infrastructure, Green Spectrum Manage- was presented with a focus on emerging IT technologies.
ment, Green Communication and Green Security and Cloud computing, mobile computing, big data analytics,
Quality of Service (QoS) Provisioning [37]: IoT, and software-based networks were identified as the
emerging IT technologies driving the current popularity
of the IT industry. The demand and social integration
• Green IoT Architectures: IoT architecture is still
of IT technologies is increasing rapidly, hence, increas-
under standardization. The committees of
ing the energy consumption. With a renewed focus on
standardization are trying to enable communication
the global energy crisis, IT researchers and practition-
between heterogeneous networks, containing various
ers have proposed and implemented several algorithms
types of devices, across various applications. The
and protocols for the green operation of the IT industry.
challenge is that communication protocols and
These algorithms and protocols implement mechanisms
devices should also consider energy-efficiency while
such as idle sleep states, energy-aware decision making,
performing their duties as anticipated by end users.
and resource scheduling. However, minimizing the energy
• Green infrastructure: Providing energy-efficient
consumption of a system significantly affects its perfor-
infrastructure for IoT is considered an important
mance parameters. The energy optimization level for a
issue towards greening. Green infrastructure can be
device highly depends on the use case of the applica-
achieved through a clean-slate redesign approach.
tion. Aggressive energy minimization policies effect sys-
Redeploying and adapting existing infrastructure is a
tem durability due to frequent power off and on system
complex task.
routines.
• Green Communication: Communication is one of the
An overall analysis of the state-of-the-art in green com-
influential factors in greening IoTs. Energy efficient
puting shows that the green algorithms and protocols are
communication between IoT nodes faces several
reaching a high level of maturity, and significant efficien-
challenges, such as supporting energy-efficient
cies are possible. In contrast, the study has demonstrated
communication protocols along with reliable
that, in the IT industry, governance is lagging signifi-
connectivity.
cantly behind, and hence consideration of green practices
• Green Security and QoS Provisioning:
is a high priority. In particular, green computing prac-
Implementation of reliable security and privacy
tices need to be implemented at the organizational level
algorithms puts the burden of computation on IoT
to complement and enforce the underlying optimization
devices, consequently it increases the energy
techniques and technologies proposed by researchers.
consumption.
The strength of green computing solutions lies in their
diversity, with consideration of low-level processor, mem-
Computing architectures, circuits, protocols, and algo- ory, and network components for system optimization
rithms are advancing innovations on green challenges alongside greedy and evolutionary heuristics. However,
faced by IT. Similarly, the efficiencies of the energy sys- again, this must coincide with robust and intelligent
tems have also shown reasonable growth over the last strategies that consider the overall performance energy
decade. The demand and popularity for computing sys- trade-offs in terms of multi-objective optimization. The
tems, storage devices, and networks has also increased, paper highlights that further research is required to ana-
hence, neutralizing the advances in green computing. lyze the impact of energy optimization techniques on
While researchers recognize the importance of contin- system performance parameters such as throughput, and
ued innovations in efficient and sustainable computing response time. This analysis of system performance and
and energy systems, industrial practices lag behind in energy will lead to more fine-tuned solutions for green
the adoption of green computing. Operational costs of computing that will be more acceptable to IT industry
computing systems can significantly decrease on adop- governors who prioritize performance parameters rather
tion of green computing practices benefiting both service than energy.
consumers and managers. IT enabled businesses and
industries need to comprehend the advantages of green Abbreviations
computing in terms of customer value, operational ATS: Autonomic transfer switch; CDC: Cloud data center; CMOS:
Complementary metal oxide semiconductor; CRT: Cathode ray tube; DFS:
cost sustainability, and environmental sustainability. The Dynamic frequency scaling; GHG: Green house gases; IoT: Internet of things;
future of green computing lies in effective endorsement IT: Information technology; M2M: Machine-to-machine; QoS: Quality of service;
Shuja et al. Journal of Internet Services and Applications (2017) 8:9 Page 10 of 11

RFID: Radio frequency identification; SOC: State of charge; VM: Virtual machine; 12. Kołodziej J, Khan SU, Wang L, Byrski A, Min-Allah N, Madani SA.
WSN: Wireless sensor network Hierarchical genetic-based grid scheduling with energy optimization.
Clust Comput. 2013;16(3):591–609.
Acknowledgements 13. Xiao Z, Song W, Chen Q. Dynamic resource allocation using virtual
The authors thank the anonymous reviewers for helpful comments that machines for cloud computing environment. IEEE Trans Parallel Distrib
increased the quality of the article. Syst. 2013;24(6):1107–17.
14. Liu H, Jin H, Xu CZ, Liao X. Performance and energy modeling for live
Funding migration of virtual machines. Clust Comput. 2013;16(2):249–64.
This work is partially funded by the Malaysian Ministry of Education under the 15. Tso FP, Hamilton G, Oikonomou K, Pezaros DP. Implementing scalable,
High Impact Research Grant of University of Malaya network-aware virtual machine migration for cloud data centers. In: IEEE
UM.C/625/1/HIR/MOE/FCSIT/03. CLOUD; 2013. p. 557–564.
16. Kokkinos P, Kalogeras D, Levin A, Varvarigos E. Survey: Live migration
Authors’ contributions
and disaster recovery over long-distance networks. ACM Comput Surv
JS contributed towards the sections of Introduction, Green Cloud Computing,
(CSUR). 2016;49(2):26.
and Conclusion. RW wrote the section Green Mobile Computing. AI wrote the
17. Goiri Í, Haque ME, Le K, Beauchea R, Nguyen TD, Guitart J, Torres J,
section Green IoT. AS contributed towards section Green Big Data Analytics.
Bianchini R. Matching renewable energy supply and demand in green
KN contributed to Green Practices and Green Computing Requirements. AG,
datacenters. Ad Hoc Netw. 2015;25:520–34.
AZ, and SK revised the manuscript. In addition, all authors read and approved
18. Markovic DS, Zivkovic D, Branovic I, Popovic R, Cvetkovic D. Smart
the work.
power grid and cloud computing. Renew Sust Energ Rev. 2013;24:566–77.
Competing interests 19. Schomaker G, Janacek S, Schlitt D. The Energy Demand of Data Centers.
The authors declare that they have no competing interests. 2015;113–124.
20. Huang F, Lu J, Zheng J, Baleynaud J. Feasibility of heat recovery for
Publisher’s Note district heating based on cloud computing industrial park. In: 2015
Springer Nature remains neutral with regard to jurisdictional claims in International Conference on Renewable Energy Research and
published maps and institutional affiliations. Applications (ICRERA). IEEE; 2015. p. 287–291.
21. Jing SY, Ali S, She K, Zhong Y. State-of-the-art research study for green
Author details cloud computing. J Supercomput. 2013;65(1):445–68.
1 COMSATS Institute of Information Technology, University Road, Abbottabad,
22. Juarez F, Ejarque J, Badia RM. Dynamic energy-aware scheduling for
Pakistan. 2 Centre for Mobile Cloud Computing Research (C4MCCR), FSKTM, parallel task-based application in cloud computing. Future Generation
University Malaya, Kuala Lumpur, Malaysia. 3 Faculty of Computing and Computer Systems. 2016. ISSN 0167-739X. http://dx.doi.org/10.1016/j.
Informatics, University Malaysia Sabah, Kota Kinabalu, Malaysia. 4 Department future.2016.06.029.
of Electrical and Computer Engineering, North Dakota State University, Fargo, 23. Li X, Ortiz PJ, Browne J, Franklin D, Oliver JY, Geyer R, Zhou Y, Chong
ND 58108-6050, Fargo, USA. 5 School of Information Technologies, University FT. Smartphone evolution and reuse: Establishing a more sustainable
of Sydney, Sydney, Australia. model. In: 2010 39th International Conference on Parallel Processing
Workshops. IEEE; 2010. p. 476–84.
Received: 31 January 2017 Accepted: 28 June 2017 24. Ahmad RW, Gani A, Hamid SHA, Xia F, Shiraz M. A review on mobile
application energy profiling: Taxonomy, state-of-the-art, and open
research issues. J Netw Comput Appl. 2015;58:42–59.
References 25. Kumar BPV, Sharma NM, Kishore KL, Goel N. Leakage power recovery in
1. Shuja J, Gani A, Shamshirband S, Ahmad RW, Bilal K. Sustainable cloud spare cells by using state dependent leakage tables from library models.
data centers: A survey of enabling techniques and technologies. Renew In: 2012 Asia Pacific Conference on Postgraduate Research in
Sust Energ Rev. 2016;62:195–214. Microelectronics and Electronics. IEEE; 2012. p. 19–24.
2. Wang X, Vasilakos AV, Chen M, Liu Y, Kwon TT. A survey of green mobile 26. Blume H, Becker D, Rotenberg L, Botteck M, Brakensiek J, Noll TG.
networks: Opportunities and challenges. Mob Netw Appl. 2012;17(1):4–20. Hybrid functional-and instruction-level power modeling for embedded
3. Abbasi Z, Jonas M, Banerjee A, Gupta S, Varsamopoulos G. Evolutionary and heterogeneous processor architectures. J Syst Archit. 2007;53(10):
Green Computing Solutions for Distributed Cyber Physical Systems. 689–702.
Berlin, Heidelberg: Springer Berlin Heidelberg; 2013, pp. 1–28. 27. Shuja J, Gani A, Bilal K, Khan AUR, Madani SA, Khan SU, Zomaya AY. A
4. Deng W, Liu F, Jin H, Li B, Li D. Harnessing renewable energy in cloud survey of mobile device virtualization: Taxonomy and state of the art.
datacenters: opportunities and challenges. IEEE Netw. 2014;28(1):48–55. ACM Comput Surv (CSUR). 2016;49(1):1.
5. Zhou J, Sun J, Athukorala K, Wijekoon D, Ylianttila M. Pervasive social 28. Fernando N, Loke SW, Rahayu W. Mobile cloud computing: A survey.
computing: augmenting five facets of human intelligence. J Ambient Futur Gener Comput Syst. 2013;29(1):84–106.
Intell Humanized Comput. 2012;3(2):153–66. 29. Liu Y, Xu C, Cheung SC. Diagnosing energy efficiency and performance
6. Ahmed E, Yaqoob I, Gani A, Imran M, Guizani M. for mobile internetware applications. IEEE Soft. 2015;32(1):67–75.
Internet-of-things-based smart environments: state of the art, taxonomy, 30. Hoque MA, Siekkinen M, Khan KN, Xiao Y, Tarkoma S. Modeling,
and open research challenges. IEEE Wirel Commun. 2016;23(5):10–6. profiling, and debugging the energy consumption of mobile devices.
7. Tapia AH, LaLone NJ, Kim HW. Run amok: group crowd participation in ACM Comput Surv (CSUR). 2016;48(3):39.
identifying the bomb and bomber from the Boston marathon bombing. 31. Hardy D, Puaut I. Static probabilistic worst case execution time
In: proceedings of the 11th International ISCRAM Conference. estimation for architectures with faulty instruction caches. Real-Time Sys.
Pennsylvania; 2014. 2015;51(2):128–52.
8. Lund D, MacGillivray C, Turner V, Morales M. Worldwide and regional 32. Pananilath I, Acharya A, Vasista V, Bondhugula U. An optimizing code
internet of things (iot) 2014–2020 forecast: A virtuous circle of proven generator for a class of lattice-boltzmann computations. ACM Trans
value and demand. International Data Corporation (IDC), Tech Rep. 2014. Archit Code Optim (TACO). 2015;12(2):14.
9. Gens F, Adam M, Bradshaw D, et al. Worldwide and regional public it 33. Luporini F, Varbanescu AL, Rathgeber F, Bercea GT, Ramanujam J, Ham
cloud services 2014–2018 forecast: IDC Corporate USA; 2014. DA, Kelly PH. Cross-loop optimization of arithmetic intensity for finite
10. Shuja J, Bilal K, Madani SA, Othman M, Ranjan R, Balaji P, Khan SU. element local assembly. ACM Trans Archit Code Optim (TACO).
Survey of techniques and architectures for designing energy-efficient 2015;11(4):57.
data centers. IEEE Syst J. 2016;10(2):507–19. 34. Ba H, Heinzelman W, Janssen CA, Shi J. Mobile computing-a green
11. Ebrahimi K, Jones GF, Fleischer AS. A review of data center cooling computing resource. In: Wireless Communications and Networking
technology, operating conditions and the corresponding low-grade Conference (WCNC) 2013 IEEE. IEEE; 2013. p. 4451–6.
waste heat recovery opportunities. Renew Sust Energ Rev. 2014;31: 35. Murugesan S. Harnessing green it: Principles and practices. IT Prof.
622–38. 2008;10(1):24–33.
Shuja et al. Journal of Internet Services and Applications (2017) 8:9 Page 11 of 11

36. Juels A. Rfid security and privacy: a research survey. IEEE J Sel Areas 59. Mijumbi R, Serrat J, Gorricho JL, Bouten N, De Turck F, Boutaba R.
Commun. 2006;24(2):381–94. Network function virtualization: State-of-the-art and research challenges.
37. Shaikh FK, Zeadally S, Exposito E. Enabling technologies for green IEEE Commun Surv Tutorials. 2016;18(1):236–62.
internet of things. IEEE Syst J. 2015;PP(99):1–12. 60. Bolla R, Lombardo C, Bruschi R, Mangialardi S. Dropv2: energy efficiency
doi:10.1109/JSYST.2015.2415194. through network function virtualization. IEEE Netw. 2014;28(2):26–32.
38. Hartung C, Han R, Seielstad C, Holbrook S. Firewxnet: A multi-tiered 61. Fang C, Yu FR, Huang T, Liu J, Liu Y. A survey of green
portable wireless system for monitoring weather conditions in wildland information-centric networking: Research issues and challenges. IEEE
fire environments. In: Proceedings of the 4th international conference on Commun Surv Tutorials. 2015;17(3):1455–72.
Mobile systems, applications and services. ACM; 2006. p. 28–41. 62. Ahmad W, Hölzenspies PK, Stoelinga M, van de Pol J. Green computing:
39. Zhu C, Leung VC, Shu L, Ngai ECH. Green internet of things for smart power optimisation of vfi-based real-time multiprocessor dataflow
world. IEEE Access. 2015;3:2151–62. applications. In: Digital System Design (DSD), 2015 Euromicro Conference
40. Jeon SH, Lee J, Choi JK. An adaptive reporting frequency control scheme on. IEEE; 2015. p. 271–5.
for energy saving on wireless sensor networks. In: Information and 63. Preden J, Kaugerand J, Suurjaak E, Astapov S, Motus L, Pahtma R. Data
Communication Technology Convergence (ICTC), 2014 International to decision: pushing situational information needs to the edge of the
Conference on. IEEE; 2014. p. 39–44. network. In: 2015 IEEE International Multi-Disciplinary Conference on
41. Morreale P, Qi F, Croft P. A green wireless sensor network for Cognitive Methods in Situation Awareness and Decision. IEEE; 2015. p.
environmental monitoring and risk identification. Int J Sens Netw. 158–64.
2011;10(1-2):73–82. 64. Bolla R, Bruschi R, Davoli F, Cucchietti F. Energy efficiency in the future
42. Seah WK, Eu ZA, Tan HP. Wireless sensor networks powered by ambient internet: a survey of existing approaches and trends in energy-aware fixed
energy harvesting (wsn-heap)-survey and challenges. In: Wireless network infrastructures. IEEE Commun Surv Tutorials. 2011;13(2):223–44.
Communication, Vehicular Technology, Information Theory and 65. Liaqat M, Ninoriya S, Shuja J, Ahmad RW, Gani A. Virtual machine
Aerospace & Electronic Systems Technology, 2009. Wireless VITAE 2009. migration enabled cloud resource management: A challenging task.
1st International Conference on. IEEE; 2009. p. 1–5. 2016. arXiv preprint arXiv:160103854.
43. Gani A, Siddiqa A, Shamshirband S, Hanum F. A survey on indexing
techniques for big data: taxonomy and performance evaluation. Knowl
Inf Syst. 2016;46(2):241–84.
44. Hashem IAT, Anuar NB, Gani A, Yaqoob I, Xia F, Khan SU. Mapreduce:
Review and open challenges. Scientometrics. 20161–34.
45. Baillie R. Big data goes green. 2012. URL http://www.
renewableenergyworld.com/articles/print/volume-15/issue-3/solar-
energy/big-data-goes-green.html. Accessed 11 Jan 2017.
46. Patra C, Nath A. Green computing–new paradigm of energy efficiency
and e-waste minimization–a pilot study on current trends. Int J Adv Res
Comput Sci Manag Stud. 2014;2(11):
47. Waas FM. Beyond conventional data warehousin - massively parallel data
processing with greenplum database. In: International Workshop on
Business Intelligence for the Real-Time Enterprise. Springer; 2008. p.
89–96.
48. Goiri Í, Le K, Nguyen TD, Guitart J, Torres J, Bianchini R. Greenhadoop:
leveraging green energy in data-processing frameworks. In: Proceedings
of the 7th ACM european conference on Computer Systems. ACM; 2012.
p. 57–70.
49. “Pike-Research”. Pike research: Cloud computing energy efficiency. 2012.
URL http://www.areadevelopment.com/StudiesResearchPapers/1-3-
2011/pike-research-cloud-computing-energy49922.shtml. Accessed 11
Jan 2017.
50. ‘CDP’. Re100 annual report 2016 - growing market demand for renewable
power. 2016;17.
51. Al-Jarrah OY, Yoo PD, Muhaidat S, Karagiannidis GK, Taha K. Efficient
machine learning for big data: A review. Big Data Res. 2015;2(3):87–93.
52. Bourne PE, Lorsch JR, Green ED. Perspective: Sustaining the big-data
ecosystem. Nature. 2015;527(7576):S16—S17.
53. Asad Z, Chaudhry MAR. A two-way street: Green big data processing for a
greener smart grid. IEEE Syst J. 2016;PP(99):1–11.
doi:10.1109/JSYST.2015.2498639.
54. Shuja J, Bilal K, Madani S, Khan S. Data center energy efficient resource
scheduling. Clust Comput. 2014;17(4):1265–77.
doi:10.1007/s10586-014-0365-0.
55. Bilal K, Malik SUR, Khalid O, Hameed A, Alvarez E, Wijaysekara V, Irfan R,
Shrestha S, Dwivedy D, Ali M, et al. A taxonomy and survey on green
data center networks. Futur Gener Comput Syst. 2014;36:189–208.
56. Xia W, Wen Y, Foh CH, Niyato D, Xie H. A survey on software-defined
networking. IEEE Commun Surv Tutorials. 2015;17(1):27–51.
57. Heller B, Seetharaman S, Mahadevan P, Yiakoumis Y, Sharma P,
Banerjee S, McKeown N. Elastictree: Saving energy in data center
networks. In: Nsdi, vol 10. 2010. p. 249–64.
58. Rawat DB, Reddy SR. Software defined networking architecture, security
and energy efficiency: A survey. IEEE Commun Surv Tutorials. 2016;19.1:
325-346.