Limits...
Ubiquitous green computing techniques for high demand applications in Smart environments.

Zapater M, Sanchez C, Ayala JL, Moya JM, Risco-Martín JL - Sensors (Basel) (2012)

Bottom Line: Ubiquitous sensor network deployments, such as the ones found in Smart cities and Ambient intelligence applications, require constantly increasing high computational demands in order to process data and offer services to users.This paper proposes an energy-minimization workload assignment technique, based on heterogeneity and application-awareness, that redistributes low-demand computational tasks from high-performance facilities to idle nodes with low and medium resources in the WSN infrastructure.These non-optimal allocation policies reduce the energy consumed by the whole infrastructure and the total execution time.

View Article: PubMed Central - PubMed

Affiliation: CEI Campus Moncloa, UCM-UPM, Madrid 28040, Spain. marina@die.upm.es

ABSTRACT
Ubiquitous sensor network deployments, such as the ones found in Smart cities and Ambient intelligence applications, require constantly increasing high computational demands in order to process data and offer services to users. The nature of these applications imply the usage of data centers. Research has paid much attention to the energy consumption of the sensor nodes in WSNs infrastructures. However, supercomputing facilities are the ones presenting a higher economic and environmental impact due to their very high power consumption. The latter problem, however, has been disregarded in the field of smart environment services. This paper proposes an energy-minimization workload assignment technique, based on heterogeneity and application-awareness, that redistributes low-demand computational tasks from high-performance facilities to idle nodes with low and medium resources in the WSN infrastructure. These non-optimal allocation policies reduce the energy consumed by the whole infrastructure and the total execution time.

Show MeSH
Execution time (in seconds) for tasks in their class.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3472849&req=5

f9-sensors-12-10659: Execution time (in seconds) for tasks in their class.

Mentions: Task classification of the first and the subsequent job-sets: using all the characteristics obtained during the profiling step, a naive k-means algorithm splits the different tasks into three different classes, according to their computational demands. A projection of the resulting clustering on the energy and time axis is shown in Figure 8. As expected, the low-energy tasks (which also have low CPU-demand) are assigned to low-demand classes and the CPU-intensive tasks are divided into mid-demand and high-demand classes. According to this clustering and because all the tasks of the job set are labelled, each task will be automatically assigned to one of the classes. As in this paper we are trying to assign low-demand tasks to low-resource nodes, mid-demand tasks to midresource nodes and high-demand tasks to the HPC facility by means of the allocation algorithm, a good clustering will be one that splits tasks such that their execution time in each processor is coherent; that is, the allocated task properly adapts to the resources it has been assigned to. In order to validate our clustering, we execute the tasks in the processors where they are classified and we measure their execution time. Results are shown in Figure 9. As it can be seen, tasks are executed within proper time limits, and there are not low-resource tasks that need too much execution time. Because the Base Stations consume less power than the Gateways, and the Gateways less than the HPC servers, we can also conclude that the energy graphic will have the same shape as the time graphic of Figure 9.


Ubiquitous green computing techniques for high demand applications in Smart environments.

Zapater M, Sanchez C, Ayala JL, Moya JM, Risco-Martín JL - Sensors (Basel) (2012)

Execution time (in seconds) for tasks in their class.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3472849&req=5

f9-sensors-12-10659: Execution time (in seconds) for tasks in their class.
Mentions: Task classification of the first and the subsequent job-sets: using all the characteristics obtained during the profiling step, a naive k-means algorithm splits the different tasks into three different classes, according to their computational demands. A projection of the resulting clustering on the energy and time axis is shown in Figure 8. As expected, the low-energy tasks (which also have low CPU-demand) are assigned to low-demand classes and the CPU-intensive tasks are divided into mid-demand and high-demand classes. According to this clustering and because all the tasks of the job set are labelled, each task will be automatically assigned to one of the classes. As in this paper we are trying to assign low-demand tasks to low-resource nodes, mid-demand tasks to midresource nodes and high-demand tasks to the HPC facility by means of the allocation algorithm, a good clustering will be one that splits tasks such that their execution time in each processor is coherent; that is, the allocated task properly adapts to the resources it has been assigned to. In order to validate our clustering, we execute the tasks in the processors where they are classified and we measure their execution time. Results are shown in Figure 9. As it can be seen, tasks are executed within proper time limits, and there are not low-resource tasks that need too much execution time. Because the Base Stations consume less power than the Gateways, and the Gateways less than the HPC servers, we can also conclude that the energy graphic will have the same shape as the time graphic of Figure 9.

Bottom Line: Ubiquitous sensor network deployments, such as the ones found in Smart cities and Ambient intelligence applications, require constantly increasing high computational demands in order to process data and offer services to users.This paper proposes an energy-minimization workload assignment technique, based on heterogeneity and application-awareness, that redistributes low-demand computational tasks from high-performance facilities to idle nodes with low and medium resources in the WSN infrastructure.These non-optimal allocation policies reduce the energy consumed by the whole infrastructure and the total execution time.

View Article: PubMed Central - PubMed

Affiliation: CEI Campus Moncloa, UCM-UPM, Madrid 28040, Spain. marina@die.upm.es

ABSTRACT
Ubiquitous sensor network deployments, such as the ones found in Smart cities and Ambient intelligence applications, require constantly increasing high computational demands in order to process data and offer services to users. The nature of these applications imply the usage of data centers. Research has paid much attention to the energy consumption of the sensor nodes in WSNs infrastructures. However, supercomputing facilities are the ones presenting a higher economic and environmental impact due to their very high power consumption. The latter problem, however, has been disregarded in the field of smart environment services. This paper proposes an energy-minimization workload assignment technique, based on heterogeneity and application-awareness, that redistributes low-demand computational tasks from high-performance facilities to idle nodes with low and medium resources in the WSN infrastructure. These non-optimal allocation policies reduce the energy consumed by the whole infrastructure and the total execution time.

Show MeSH