Cloud computing is an important paradigm in Information Technology field. In Data center, where all physical resources are available, machine consumes power and emits heat which affects the environmental conditions. The Green Cloud computing solves the problem of global warming by providing eco-friendly environment. We studied that the heat emission increases with increase in energy consumption. The main aim of Green Cloud computing is to reduce the energy consumed by physical resources in data center and save energy and also increases the performance of the system. In this paper, we describe various techniques for minimizing the energy consumption.
|Green cloud computing, Virtualization, Energy consumption, Job scheduling.
|Cloud computing is internet based computing which provides web services through service providers . These
services are provided to the users on rent like pay-as-use model in which user have to pay according to the access or
use of the services. The three basic types of services are provided: Infrastructure as a service (Iaas), Platform as a
service (Paas) and Software as a service (Saas). Iaas provides the physical resources such as memory, processor etc.
Paas provides the framework or platform for developing their own applications by using cloud and there is no need to
install any platform on their own machine. Paas services such as .Net etc. Saas is basically used for running the e xisting
applications like facebook. The user does not deal with installation of any software on their physical machine. The
cloud provides such software for running these types of applications.
|In cloud, the machines are running for providing web services and these machines also consumes some
amount of energy for working. The cloud computing which focuses on reduction of energy consumption is known as
Green Cloud computing . In data center, the physical machines emit heat and harmful gases. The green cloud
computing can also be used for e-waste management . The reduction of energy consumption can be controlled on
two basis: one is hardware and other is software. For controlling energy consumption on hardware basis, the hardware
devices are used and likewise on software devices by using program and algorithms. The energy consumption
controlled at software level is easy to maintain and less expensive. These techniques have to be implemented only once
and used for data centers. In this paper we are using algorithm for minimizing the energy consumption.
|Cloud has virtual machines and machine has number of jobs for execution. There is a need for proper job
scheduling because if there are three processes such as P1, P2 and P3. The amount of energy consumed by process P2
is less than the amount of energy consumed by P1 and the amount of energy consumed by P3 is less than P2 but if we
follow FCFS (First Come First Serve) a lgorith m then P2 can’t execute before completion of P1 and P3 can’t execute
before completion of P2. This method consumes much more energy when waiting. We have to reduce the waiting
II. RELATED WORK
|In data centers, if the physical machines consume more energy, these resources will also emit heat and harmful
gases. This problem can be minimized by Green Cloud computing which provides the eco -friendly environment. The
Green cloud computing reduces the energy consumption and also save energy.
|A. Virtualization and Cooling technique
|Authors in  provide another solution for greening the data centers. First is cooling system which minimizes
the energy consumption. Various companies uses river water for data center’s cooling, open air data centers, air -
conditioned system etc. This system is expensive and not efficient for minimizing energy consumption. Second is
virtualizat ion technique in which more than one virtual machine loaded on a single physical machine. The virtualization
technique provides the abstraction because the internal working hidden from the user. The user only accesses the web
services and does not aware about the virtual machines. The virtualization techniques realize that a single physical
machine is provided to the single user. This reduces the energy consumption because of single physical machine
running. But performance can be degraded, the reason behind the performance degradation is a unbalanced load. Third
is nano data center technique which specifies that the large number of small sized data centers should be geographical ly
distributed. Traditional data centers are of large size and few data centers are distributed and this technique consumes
more energy because of long distance of data transmission. This nano data center technique reduces the energy
consumption because of the short distance between client and data center. These all techniques are used for greening
the data centers.
|B. Energy management in public and private cloud
|In cloud computing there are two basic clouds first is public cloud and second is private cloud. Public cloud is
accessible from any user through internet but private cloud is only accessible by the particular organization. The
analysis of energy consumption is performed on basic web services such as storage as a service, software as a service
and processing as a service . Storage as a service provides a service in which user can store their data on cloud not
on their personal machine. There is no need to buy any storage device such as hard disk, but the user have to pay
according to the usage of the storage devices on cloud. Software as a service provides the latest software to the user
through cloud for developing their own applications easily. There is no need to get license for software. Processing as a
service is used for performing the computations on user’s data and after all operations the result is provided to the user.
|There are various energy consumption models which consumes energy. First, user equipments such as
processor, memory, display unit etc. these devices consumes energy but at us er side. Second, data center consumes
energy because there are number of devices used for providing the services to the users. The energy consumption can
be reduced by consolidating the servers but for consolidation the servers which are idle and have no t ask to perform can
be turned off. In this method the load is distributed to the few servers and performance can be degraded. This process
requires more attention.
|The energy consumption analysis is performed on three web services such as storage as a service, software as
a service and processing as a service. In case of storage as a service, the user creates their file and store on cloud. After
some time if user wants to edit this file then the user must download the file form cloud and after providing the
modifications again upload the file on cloud. This process consumes more energy because of uploading and
downloading the files on cloud.
|Fig.1.shows the comparison of public cloud and private cloud in which the public cloud consumes more
energy than private cloud because of load on cloud. The private cloud is accessed by only the members of the
organization but the public cloud is accessed by any user. In case of software as a service it also consumes energy for
transporting the frame work on user’s machine through terminal. Last service is processing as a service consumes more
energy in public cloud than a private cloud.
|C. Job Scheduling
|In , scheduler schedules the tasks by determining the temperature of the task and node. The tasks are
generated by Task Generation System. This system determines the temperature of the task by specifying the parameters
such as initial temperature of the task, per minute rise in temperature and execution time of the task. This specification
is given manually.
|After determining the temperature of the task then the prediction method is used for determining the
temperature of the node. This prediction method uses two parameters: 1) task specification and 2) energy consumption.
In this scheduler FCFS (First Come First Serve) algorithm and priority algorithm is used for scheduling. The priority
algorithm schedules tasks according to the temperature of the task and node.
|The task, which requires low temperature, has high priority and the high temperature task have low priority. In
this algorithm, one additional parameter is used for comparison which is a critical temperature and if any task requires
temperature up to critical temperature then this task will not be executed, otherwise system gets failure. Fig.2. shows
comparison chart with scheduler and without scheduler. Author says that system performance should not be impacted
while energy consumption is being minimized.
|Power aware virtual machine scheduling is another technique for reduction of energy cons umption . The
virtual machines are scheduled according the power consumed by the virtual machines. This scheduling is provided for
minimizing the performance overheads but with energy efficiency. But this technique does not providing the greenest data center which
is the main aim of green cloud computing. The job grouping is another technique for efficient energy consumption .
Jobs are scheduled according to the resource capability. Before the scheduling process, calculate the capability of each
resource by selecting them. After calculating the capability resources then allocate the jobs to the resources according
their capability. This scheduling technique is basically used for load balancing but with minimum reduction of energy
consumption. In , the jobs are grouped together on the basis of similar resource requirement. This scheduling
technique concentrates only on efficient resource management but with minimal reduction of energy consumption. This
reduction is provided by reducing the waiting energy of the jobs.
|In this paper, we present various techniques for reducing the energy consumption by resources in data centers. The
Green cloud computing can achieve this goal. Virtualization technique mainly reduces the energy consumption but
increases the performance overhead because a single physical machine is responsible for managing various virtual
machines. The cooling technique can provide the greenest computing technology but it is an expensive technique.
Another technique for reducing the energy consumption is job scheduling and it is not a part of hardware system. This
technique is better than other techniques. The existing job s cheduling algorithms based on energy consumed by the
virtual machines & capability of resources. The priority algorithms schedule jobs according the temperature of the node
and groups of jobs. The priority algorithm with grouping is better than other algorithms because it increases the overall
performance. Research on this field is very important for environmental conditions.
Figures at a glance
- R. Buyya, A. Beloglazov, J. Abawajy, ?Energy-aware resource allocat ion heurist ic for efficient management of dat a cent ers for cloud comput ing?, Future Generat ion Computer Systems, pp.755-768, 2012.
- B. Agrawal, ? Green Cloud Comput ing?, Int ernat ional Journal of Elect ronics and Communicat ion Engineering & T echnology?, pp. 239 -243, 2013.
- D. Cavdar, F. Alagoz, ? A Survaey of Research on Greening Dat a Centers?, IEEE Symposium on Select ed Areas in Communicat ions, pp. 3237-3242, 2012.
- J. Baliga, R. Ayre, K. Hint on, R. T ucker, ? Green Cloud Comput ing: Balancing Energy in Processing, Storage and Transport?, P roceedings of the IEEE, Vol. 99, pp. 149-167, 2011.
- S. Arora, V. Chopra, ? A Predict ive Energy Aware Hybrid Resource Scheduler for Green Cloud Comput ing?, Int ernat ional Journal of Applied Research in Comput ing, Vol. 1, pp. 1-5, 2013.
- Y. P onnusamy, S. Sasikumar, ?Application of Green Cloud Comput ing for Energy Management?, Internat ional Joural of Comput er Science Informat ion Technology, Vol. 3 (5), pp. 5101-5107, 2012.
- S. Selvarani, ?Improved Job-Grouping based P SO algorithm for T ask scheduling?, Int ernat ional Journal of Engineering Science and T echnology, Vol. 2(9), pp. 4687-4695, 2010.
- R. Nanduri, N. Maheshwari, R. Raja, V. Varma, ? Job Aware Scheduling Algorit hm for MapReduce Framework?, IEEE Internat ional Conference on Cloud Comput ing Technology and Science, pp. 3-4, 2011.