Limits...
A Novel Two-Tier Cooperative Caching Mechanism for the Optimization of Multi-Attribute Periodic Queries in Wireless Sensor Networks.

Zhou Z, Zhao D, Shu L, Tsang KF - Sensors (Basel) (2015)

Bottom Line: Usually, certain sensory data may not vary significantly within a certain time duration for certain applications.Leveraging these cooperatively cached sensory data, queries are answered through composing these two-tier cached data.Experimental evaluation shows that this approach can reduce the network communication cost significantly and increase the network capability.

View Article: PubMed Central - PubMed

Affiliation: School of Information Engineering, China University of Geosciences (Beijing), Beijing 100083, China. zhangbing.zhou@gmail.com.

ABSTRACT
Wireless sensor networks, serving as an important interface between physical environments and computational systems, have been used extensively for supporting domain applications, where multiple-attribute sensory data are queried from the network continuously and periodically. Usually, certain sensory data may not vary significantly within a certain time duration for certain applications. In this setting, sensory data gathered at a certain time slot can be used for answering concurrent queries and may be reused for answering the forthcoming queries when the variation of these data is within a certain threshold. To address this challenge, a popularity-based cooperative caching mechanism is proposed in this article, where the popularity of sensory data is calculated according to the queries issued in recent time slots. This popularity reflects the possibility that sensory data are interested in the forthcoming queries. Generally, sensory data with the highest popularity are cached at the sink node, while sensory data that may not be interested in the forthcoming queries are cached in the head nodes of divided grid cells. Leveraging these cooperatively cached sensory data, queries are answered through composing these two-tier cached data. Experimental evaluation shows that this approach can reduce the network communication cost significantly and increase the network capability.

No MeSH data available.


Comparison of cache hit rates for MAQWR, where the number of attributes are set to 1, 3, 5, 7 and 9, respectively. Similar to Figure 4, when the cache size of the SN is relatively small and is not capable of caching all sensory data requested by a certain query, the cache hit rates for sensory data cached in the SN decrease significantly (roughly from 95% down to 70%).
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4541820&req=5

f8-sensors-15-15033: Comparison of cache hit rates for MAQWR, where the number of attributes are set to 1, 3, 5, 7 and 9, respectively. Similar to Figure 4, when the cache size of the SN is relatively small and is not capable of caching all sensory data requested by a certain query, the cache hit rates for sensory data cached in the SN decrease significantly (roughly from 95% down to 70%).

Mentions: Figure 8 shows cache hit rates hrtcah for MAQWR. hrtcah is calculated as the ratio of , where (i) is the set of sensory data cached in the SN that contributes to answering of the query q and (ii) SDq is the set of sensory data of q inquiries. Without loss of generality, the value of sensory data is assumed to vary according to the formula: valvsn = log (k × tcur + 1) + C, where: (i) tcur is the current time; and (ii) k and C are constants, which are initially set to random values, and vary according to a normal distribution. Therefore, sensory data of sensor nodes are mostly different and change moderately. Figure 8 shows that cache hit rates for the scenarios, where the number of attributes is 1, 3 or 5, are quite high (roughly 95%), since the SN can have enough storage capability to cache almost all sensory data gathered in recent time slots. As for the scenarios where the number of attributes is seven or nine, cache hit rates are relatively lower (roughly 70%). Similar to the situation for energy consumption, a certain amount of sensory data has to be gathered from the network for query answering in real time. In addition, sensory data, which may be reused for answering the forthcoming queries, have to be removed from the cache by the data replacement mechanism, due to the limitation of the storage capability of the SN. Figure 8 shows that cache hit rates drop every 5 min. Since INs synchronize with sensor nodes in the corresponding grid cells every 5 min, sensory data, which have been changed remarkably, are retrieved from the network. These variations are routed to the SN, but these sensory data are not counted in , which induces the dropping of hrtcah consequently.


A Novel Two-Tier Cooperative Caching Mechanism for the Optimization of Multi-Attribute Periodic Queries in Wireless Sensor Networks.

Zhou Z, Zhao D, Shu L, Tsang KF - Sensors (Basel) (2015)

Comparison of cache hit rates for MAQWR, where the number of attributes are set to 1, 3, 5, 7 and 9, respectively. Similar to Figure 4, when the cache size of the SN is relatively small and is not capable of caching all sensory data requested by a certain query, the cache hit rates for sensory data cached in the SN decrease significantly (roughly from 95% down to 70%).
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4541820&req=5

f8-sensors-15-15033: Comparison of cache hit rates for MAQWR, where the number of attributes are set to 1, 3, 5, 7 and 9, respectively. Similar to Figure 4, when the cache size of the SN is relatively small and is not capable of caching all sensory data requested by a certain query, the cache hit rates for sensory data cached in the SN decrease significantly (roughly from 95% down to 70%).
Mentions: Figure 8 shows cache hit rates hrtcah for MAQWR. hrtcah is calculated as the ratio of , where (i) is the set of sensory data cached in the SN that contributes to answering of the query q and (ii) SDq is the set of sensory data of q inquiries. Without loss of generality, the value of sensory data is assumed to vary according to the formula: valvsn = log (k × tcur + 1) + C, where: (i) tcur is the current time; and (ii) k and C are constants, which are initially set to random values, and vary according to a normal distribution. Therefore, sensory data of sensor nodes are mostly different and change moderately. Figure 8 shows that cache hit rates for the scenarios, where the number of attributes is 1, 3 or 5, are quite high (roughly 95%), since the SN can have enough storage capability to cache almost all sensory data gathered in recent time slots. As for the scenarios where the number of attributes is seven or nine, cache hit rates are relatively lower (roughly 70%). Similar to the situation for energy consumption, a certain amount of sensory data has to be gathered from the network for query answering in real time. In addition, sensory data, which may be reused for answering the forthcoming queries, have to be removed from the cache by the data replacement mechanism, due to the limitation of the storage capability of the SN. Figure 8 shows that cache hit rates drop every 5 min. Since INs synchronize with sensor nodes in the corresponding grid cells every 5 min, sensory data, which have been changed remarkably, are retrieved from the network. These variations are routed to the SN, but these sensory data are not counted in , which induces the dropping of hrtcah consequently.

Bottom Line: Usually, certain sensory data may not vary significantly within a certain time duration for certain applications.Leveraging these cooperatively cached sensory data, queries are answered through composing these two-tier cached data.Experimental evaluation shows that this approach can reduce the network communication cost significantly and increase the network capability.

View Article: PubMed Central - PubMed

Affiliation: School of Information Engineering, China University of Geosciences (Beijing), Beijing 100083, China. zhangbing.zhou@gmail.com.

ABSTRACT
Wireless sensor networks, serving as an important interface between physical environments and computational systems, have been used extensively for supporting domain applications, where multiple-attribute sensory data are queried from the network continuously and periodically. Usually, certain sensory data may not vary significantly within a certain time duration for certain applications. In this setting, sensory data gathered at a certain time slot can be used for answering concurrent queries and may be reused for answering the forthcoming queries when the variation of these data is within a certain threshold. To address this challenge, a popularity-based cooperative caching mechanism is proposed in this article, where the popularity of sensory data is calculated according to the queries issued in recent time slots. This popularity reflects the possibility that sensory data are interested in the forthcoming queries. Generally, sensory data with the highest popularity are cached at the sink node, while sensory data that may not be interested in the forthcoming queries are cached in the head nodes of divided grid cells. Leveraging these cooperatively cached sensory data, queries are answered through composing these two-tier cached data. Experimental evaluation shows that this approach can reduce the network communication cost significantly and increase the network capability.

No MeSH data available.