Sei sulla pagina 1di 6

2005 Intemational Conference on Control and Automation (ICCA2005)

June 27-29, 2005, Budapest, Hungary MP-2.5

Data Management for Wireless Sensor


Networks: Research Issues and Challenges
Li Qun Zhuang, Jing Bing Zhang, Dan Hong Zhang and Yi Zhi Zhao

developed for the purpose of network monitoring and traffic


Abstract- Wireless sensor networks (WSNs) as the basic analysis. StatStream [5] proposed efficient methods to make
infrastructure for pervasive computing environment are data- decision from tens of thousands of time series data streams
centric in nature. The research issues and chalienges relating in the financial market based on discrete Fourier transforms
to sensor data management are much different from those of
traditional database and wireless communication. This paper
and a three-level time interval hierarchy.
presents a concise review of the latest technologies, research In recent years, the research attentions on continuous
issues, and challenges facing data management for WSNs. queries have been focused on WSNs. For example,
COUGAR [6] defined sensor database and sensor queries
1. INTRODUCTION by extending the Cornell PREDATOR object-relational
THE advent of WSNs is rapidly changing the way by database system; Aurora [7] is a data-flow system that
which data are captured, processed and disseminated allows the sensor data to flow through a loop-free, directed
for a variety of potential applications for home, hospitals, graph of processing operations; the Stanford stream data
factory shop floors, road traffic, battlefields, and so on. As manager (STREAM) [8] is a general-purpose data stream
with any other new technological development, there exist management system for processing continuous data streams
many research issues and technical challenges relating to and stored relations; and TelegraphCQ [9] provided an
sensor nodes, networks and applications [1]. WSN adaptive dataflow architecture for streaming data.
applications are data-centric in nature and the sensory data WSNs are typically composed of a large number of
streams are usually characterized by append-only sequence nodes each with limited resources in terms of battery life,
for data updates, continuous queries and approximate computational power and communication capabilities.
results. Hence, many researchers treat WSNs as virtual Hence energy-aware query dissemination and routing
distributed databases of continuous data streams from techniques are required as compared to the traditional
physical world. routing technology such as flooding and gossiping [10].
Research on the modeling of data streams can be traced The emerging Web Services Resource Framework
back to early 1990's, when continuous queries were (WSRF) [11] under Open Grid Service Architecture has
developed and incorporated into the Tapestry system [2] for created great opportunities to integrate resource limited
filtering streams of electronic documents, such as mail WSN applications, which generate huge amount of data,
messages or articles. As shown in (1), the monotonous into the powerfidl Grid computation architecture that is able
query result Q,(t) denotes the total set of data retumed up to provide immense computational power.
until time t by executing query Q in the Tapestry system.
1I. DATA MODEL AND QUERY LANGUAGE
QM(t) UQ(S) (1)
s<t Most of the current WSN data stream engines proposed
In NiagaraCQ [3], group optimization was applied to by researchers can be categorized into one of the following
both time-based and change-based continuous queries for models: (1) workflow-based, (2) relation-based, or (3)
Internet database and Web contents. Gigascope [4] was object-based. A general architecture of the data stream
model for WSNs is shown in Figure 1.
This work was supported by the Agency for Science, Technology and
Research (A*STAR) Science & Engineering Research Council (SERC)
TSRP Grant 04211 20028(2004)
L. Q. Zhuang is with the Singapore Institute of Manufacturing
Technology, Singapore 638075 (phone: 65-67938993; fax: 65-67916377;
e-mail: lqzhuang@simtech.a-star.edu.sg).
J. B. Zhang, D. H Zhang, Y.Z. Zhao are with the Singapore Institute of
Manufacturing Technology, Singapore 638075 (e-mail: Figure 1: Conceptual Model for Data Stream
jbzhang(simtech.a-star.edu.sg, dzhang(simtech.a-star.edu.sg,
yzzhao~simtech.a-star.edu.sg).

208
0-7803-9137-3/05/$20.00 © 2005 IEEE
The Aurora system [7] is a workflow oriented system Rstream(R) = U(R(r) x{T}) (4)
supported by a set of operators. The system is focused on r.O
efficient scheduling, quality of service and the optimization where:R(-c) represents the time-varying relation R at timer.
structure. Aurora supports continuous queries, sliding COUGAR is an object-based data stream system [13],
windows and ad hoc queries in the unified processing which models sensor nodes as abstract data types (ADTs) in
engine. The data stream in Aurora is an append-only an object-relational or object-oriented data model. Well-
sequence of tuples with uniform type: (TS, Al, A2,. .., An) defined semantics of long-running queries over ADT
where TS is the timestamp and A- is application-specific objects are also introduced in COUGAR. Set-oriented
data field. The Aurora query is built on the Aurora query execution strategies were adapted to the distributed context
algebra including three order-agnostic operators (Filter, of a device database system. The implementation of the
Map and Union) and four order-sensitive operators (BSort, Comell COUGAR device database system is based on
Aggregate, Join and Resample). Borealis [12] extends the PREDATOR, an object relational database system [14].
Aurora architecture to a new data model with revisions: A recent data streaming architecture was proposed based
(TS, tuple-type, id, A1, A2,. , An) where tuple-type is on probabilistic model [15]. Probabilistic query is built on
supported by insertion, deletion and replacement operators probability density function p(XI, X2,.. ., XJ) to assign a
and id is the unique identifier of the tuple of the stream. probability for each possible assignment to the sensor
Borealis created a new data streaming engine with the attribute Xi. The model uses historical measurements to
following features: dynamic revision of query results, improve the confidence of query answers. It is also able to
dynamic modification of queries and flexible and highly- tolerate data loss, detect faulty inputs, and identify
scalable optimization. Both Aurora and Borealis are correlations that can be used to improve query performance.
implemented as a distributed data stream model that is Since the relational model has a strong and well defined
dynamically reconfigurable with the change of network mathematical foundation in relational algebra, relation-
conditions. based model with extension of windowing and sequence
STREAM [8], developed by Stamford University, is a operators is presently the most popular paradigm. However,
relation-based data stream manager. It is modeled as an researchers are also investigating more general stream
unbounded, append-only bag of (tuple, timestamp) pairs algebra for the data streams of WSNs.
and a relation as a time-varying bag of tuples supporting Sensor data acquisition is carried out by the query
updates, deletions and insertions. These semantics rely on processor or engine. In the query processing model for data
three abstract blocks: relation-to-relation operators, stream- collection in a sensor network, research issues arise at all
to-relation operators and relation-to-stream operators. The query processing phases: query processing, query
interactions among them are shown in Figure 2. optimization, and query dissemination. Power-based query
ts u
Rebia to Reat
Oper optimization is used to choose a query plan that will yield
the lowest overall power consumption and power-efficient
protocol is the key target for data dissemination.
III. QUERY PROCESSING AND OPTIMIZATION
Figure 2: Abstract Semantics Interaction in STREAM
Many optimization strategies for WSNs are cost-based in
Continuous Query Language (CQL) is an expressive terms of computation and communication. In the case of
SQL-based declarative language for registering continuous STREAM [8], whenever a continuous query is registered a
queries against streams and updatable relations in query plan will be generated. The query plan is composed
STREAM. It is an instantiation of abstract semantics using of three components: operators, inter-operator queues and
SQL to map from relations to relations, the sliding window synopses which maintain run-time states associated with
to map from streams to relations, and three operators to map operators. The execution of query plans is controlled by a
from relations to streams. These three operators (Insertion, global scheduler for the run-time resource management.
Deletion and Relation) are formally expressed in (2), (3) The operator scheduling strategy (known as chain
and (4) respectively. CQL has three types of sliding scheduling) is near-optimal in minimizing run-time memory
window operators: time-based, tuple-based and partitioned usage for single-stream queries. The optimization
window. The window specification language is based on techniques adopted by STREAM include relocating
SQL-99. window operators in query plans, exploiting stream
Istream(R) = U ((R(v) - R(r - 1)) x {r}) (2)
constraints to reduce window sizes and identifying
r20 opportunities for sharing computation and memory. When
Dstream(R) = U ((R(r - 1)- R(r)) x {T}) (3) resource limitations force approximate query results, it uses
r>-O techniques such as sliding window reduction, sampling rate

209
reduction, synopses summarization and graceful 0 0
degradation by load shedding.
A QoS (Quality of Service) model forms the basis of
0 o D-0 0
resource management decisions in Borealis [12]. The QoS
0
0 0
defined in a vector of metrics (VM) can be predicted at any t;ep I step 2 ° te 3

point in a data flow. In Borealis, there are three levels of 0 ~ 0


optimization: a local optimizer for scheduling message, a
neighborhood optimizer for load balancing, and a global
J4 IN4 a ON10,
optimizer for global decision making. 0 5 AY,

Another important optimization strategy is power-based.


Queries in TinyDB [16] are parsed at the base station. The Figure 3: An Example of SPIN-BC
base station stores metadata information of each sensor Directed diff-usion provides a data-centric communication
node, which describes the costs of processing and protocol [19] for sensor sources and sinks with the
delivering data, and is used in query-lifetime estimation. following key features: data-centric dissemination,
Based on the metadata, TinyDB performs reordering of the reinforcement-based adaptation to the empirically best path,
query predicate (ordering of sampling, selections and joins) and in-network data aggregation and caching. Figure 4
to optimize the query process. For event-based query and shows an example of directed diffusion.
periodic query, TinyDB employs multi-query optimization Source Sotrce ( Source
technique based on rewriting which converts extemnal
events into a stream of events, and rewrite the entire set of
independent internal queries as a sliding window join
between events and sensors. The rewriting technique is also
applied to the rate-based optimization [17] where the
Sink
i Sul*
fundamental statistics used are estimates of the rates of the Sink
Step 1: Request Diffusion Step 2: Setup Gradients Step 3: Data Transfer
streams in the query evaluation tree.
Figure 4: An Example of Directed Diffusion
IV. QUERY DISSEMINATION AND ROUTING The concept of semantic routing tree (SRT) was
The WSNs disseminate the query into the sensor proposed in TinyDB [16] to provide a power efficient
networks after query optimization. The dissemination mechanism for disseminating queries and collecting query
begins with a broadcast of the query from the root of the results. With this method, nodes in SRT choose parents
networks. Routing in a WSN protocol is different from that based on semantic properties as well as the link quality. For
in ad-hoc networks. Most of the routing protocols can be attributes that are highly correlated amongst neighbors in
classified as data-centric, cluster-based or location-based, the routing tree, SRT is able to reduce the number of nodes
with a few based on QoS. The following sections discuss involved in disseminating queries and forwarding results.
the different routing protocols for WSNs.
B. Cluster-based Routing
A. Data-centric Routing Low Energy Adaptive Clustering Hierarchy (LEACH)
The Sensor Protocol for Information via Negotiation [20] is a self-organizing, adaptive clustering protocol that
(SPIN) [18] family of protocols is iused to efficiently uses randomization to distribute the energy load evenly
disseminate information in a WSN. SPIN overcomes the among the sensors in the network. LEACH involves
shortcomings of conventional approaches by making use of randomized rotation of high energy cluster heads so that it
the metadata negotiation and resource-adaptive algorithms. does not drain out the battery of a single node. Figure 5
SPIN nodes use three types of message: ADV (data shows one routing example for LEACH. In LEACH, the
advertisement), REQ (request for data) and DATA (data operation is divided into two phases: set-up phase and
message). There are four protocols in the SPIN family: steady-state phase. During the set-up phase, each node can
1) SPIN-PP (protocol for point-to-point media, uses choose a random number between 0 and 1. If the number is
ADV-REQ-DATA messages), less than threshold T(n), this node becomes a cluster-head
2) SPIN-EC (same as SPIN-PP, but is energy aware), for the current round. The threshold is set in formula (5).
3) SPIN-BC (protocol for broadcast media, uses ADV-
REQ-DATA messages, steps are shown in Figure 3) and F if ne G
4) SPIN-RL (improved on SPIN-BC which allows T(n) = I- P* (rmod e) (5)
nodes to retransmits REQ if it doesn't receive any data after else
timeout). lo
where: P = desired percentage of cluster heads, r = the

210
current round, and G = set of nodes that have not been a C. Location-based Routing
cluster head in the last 1/P rounds. Greedy Perimeter Stateless Routing (GPSR) [25] is a
Hodlel ,- (- HNode X responsive and efficient routing protocol for WSNs, which
Hode 6
uses greedy forwarding to forward packets to nodes and
only needs local information for forwarding.
Geographic Adaptive Fidelity (GAF) [26] sets up virtual
liodle 3 grid based on location information, where each node of the
(u tlode 4 grid is equivalent peers from a traffic forwarding
Figure 5: LEACH routing protocol perspective and only one node is awake in each grid at each
time. If the average number of nodes in a grid is M, GAF
Once the cluster-heads are elected, the cluster-heads
reduces the energy by a factor M' which is shown in (6).
broadcast to all the sensor nodes in the networks that they
are the new cluster-heads. Upon receiving the broadcasting, M
M =-M (6)
sensor nodes determine the cluster to which they want to I1 -e
belong based on the signal strength of the broadcasting, and The Sleep-aWake Probabilistic Forwarding Protocol
inform the chosen cluster-heads of the willingness to be a (SW-PFR) [27] avoids flooding by favoring in a
member of the cluster. During the steady-state phase, the probabilistic way certain "close to optimal" data
sensor nodes begin to sense and transmit data to the cluster- transmissions and also allows particles to altemate between
heads. Upon aggregating data from the cluster nodes, the sleeping and awake modes to save energy.
cluster-heads send these data to the base station. After the Sequential Assignment Routing (SAR) [28] facilitates
steady-state phase last for a certain period of time, the selection of a path among multi-path by the node which
network goes into the set-up phase again and entering into generates the packet. The objective of the SAR algorithm is
another round of electing the cluster-heads. to minimize the average weighted QoS metric throughout
Threshold-sensitive Energy Efficient sensor Network the lifetime of the network.
(TEEN) [21] and Adaptive Periodic TEEN (APTEEN) [22]
protocols are follows-up of LEACH, with the introduction V. DATA VISUALIZATION AND SERVICE VIRTUALIZATION
of two thresholds: hard threshold and soft threshold. Data To integrate WSNs into the Web-based environment, two
are reported to the cluster-head only when the value of technical issues have to be addressed: the visibility and
sensed attribute is equal to or greater than the hard serviceability that a WSN application can provide. Some
threshold, or the change in the value of sensed attribute is WSN applications have been reported in areas such as
equal to or greater than the soft threshold. medical and environmental monitoring [29, 30] to address
Power-Efficient GAthering in Sensor Information these challenges.
Systems (PEGASIS) [23] is a near optimal chain-based Earlier works on the visibility of wireless sensor data
protocol, which performs better than LEACH by about were carried out by Jet Propulsion Laboratory (JPL) of
100% to 300% in terms of power-efficiency. The lifetime of NASA in the Sensor Web project [31]. The Sensor Web is a
sensor network supported by PEGASIS is almost twice as dynamic infrastructure for sensors to synthesize the
that of LEACH. Figure 6 shows one example for PEGASIS. knowledge from data, providing end-to-end solutions for
Hode-ll() odle 2 Intemet browser to access sensor networks for command in
9 J~~~~~~~~lode
6 and streaming data out.
A sensor information networking architecture (SINA)
Sink [32] was proposed to serve as the middleware to facilitate
" e f -~~Cuse Head /
querying, monitoring, and tasking of sensor networks.
EmStar [33] realized the software reusability by providing
Niode 4 Hode S module-based services that support networking, sensing,
and time synchronization and it also implemented tools that
Figure 6: An Example of PEGASIS support execution, simulation and visualization of the live
environment. The Tiny Application Sensor Kit (TASK)
The Two-Tier Data Dissemination (TTDD) [24] [34] was built on top of TinyDB to provide high level
approach provides scalable and efficient data delivery to metadata management, query configuration, sensor
multiple mobile sinks. Each data source in TTDD networks monitoring and data visualization.
proactively constructs and maintains a grid structure with Many reported WSN applications were built on
very low overhead, which enables mobile sinks to homogenous architectures. It is difficult to integrate WSNs
continuously receive data on the move by flooding queries with different devices and services, correlate between
within a local cell only. heterogeneous sensor networks and provide unified

211
services. The Stream-based Overlay Network (SBON) [35] game theory.
attempted to address the challenge of serviceability of WSN D) Integrating wireless sensor networks with the grid
applications. SBON provides large scale sensor resource poses many challenges. More challenges are foreseen to
sharing in the development of Internet-based stream by provide mechanisms to realize the automated orchestration
abstracting the details of process optimization and service and integration of services for WSNs. More research works
naming and discovery. SBON leverages recent researches are needed in the areas of ontological engineering,
in overlay networks and peer-to-peer architecture to process orchestration and choreography of WSN services. Business
and deliver stream data over the Intemet. And recently, an Process Execution Language (BPEL) is also needed to form
Internet-scale query engine called Peer-to-peer Information the semantic markup of Web Services for WSNs.
Exchange and Retrieval (PIER) [36] provides general- The research issues pertinent to data management are
purpose peer-to-peer architecture including data summarized in Table 1.
representation, query dissemination, query operators and
metadata management. TABLE 1: ISSUES OF DATA MANAGEMENT FOR WSNs: A TAXONOMY
For maximum inter-operability, WSN applications for Research Issue Reference
any specific domain should be generalized in a service- Data Modeling [6,7,8,9,12,13,14,15,16]
oriented architecture. Web services provide the fundamental
architecture for implementation. In order to manage the Query Optimization [8,12,15,16,17]
Web services, WSRF has introduced a design pattern to Routing Protocol [10,16,18,19,20,21,22,23,24,25,26,27,28]
specify how to use Web services to access "stateful" WSNs
with the following supporting modules: WS-Resource Virtualization [29,30,31,32,33,34,35,36,38]
Property, WS-Resource Lifetime, WS-Base Faults, WS-
Renewable References, WS-Service Group and WS- REFERENCES
Notification [37]. To integrate resource-constrained sensor [1] 1. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E.
nodes with the resource-rich grid computing environment, Cayirci, "Wireless sensor networks: a survey", Computer
the new network infrastructure should be built on WSRF. Networks Journal, vol. 38(4) pp.393422, 2002.
Hourglass [38] provides grid-enabled APIs for the [2] D. Terry, D. Goldberg, D. Nichols and B. Oki, "Continuous
heterogeneous sensors with WSRF. queries over append-only databases", The ACM SIGMOD of
the International Conference on Management of Data, San
Diego, USA, pp.321-330, 1992.
VI. CONCLUDING REMARKS [3] J. Chen, D.J. Dewitt, F. Tian and Y. Wang, "NiagaraCQ: A
This paper has presented a concise review of the scalable continuous query system for Internet database", The
emerging technologies for WSNs. As with any new ACM SIGMOD of International Conference on Management
of Data, Dallas, USA, pp.379-390, 2000.
technology, WSN is still at an early stage, and lacks [4] C. Cranor, T. Johnson, 0. Spatscheck, and V. Shkapenyuk,
standards in various aspects. Its immaturity, however, "Gigascope: A Stream Database for Network Applications",
creates a lot of opportunities for pioneering researchers to The ACM SIGMOD of International Conference on
innovate. As its potentials for many applications have been Management of Data, San Diego, USA, pp.647-651, 2003.
recognized, data management for WSNs has attracted a lot [5] Y. Zhu and D. Shasha, "StatStream: Statistical monitoring of
of attentions in the international communities, and many thousands of data streams in real time", The 28th
International Conference on Verv Large Data Bases, HK,
new approaches have been proposed by researchers. China, pp.358-369, 2002.
However, there are still many challenges and open issues [6] P. Bonnet, J. Gehrke and P. Seshadri, "Towards sensor
for researchers to address as the highlighted points below. database systems", The 2nd International Conference on
A) How to satisfy the constraints inherent in WSNs in Mobile Data Management, HK, China, pp.3-14, 2001.
the realization of sensor networks, such as topology change, [7] D. Abadi, D. Carney, U. Cetintemel, M. Cherniack, C.
fault tolerance, scalability, cost, hardware, power Convey, S. Lee, M. Stonebraker, N. Tatbul, and S. Zdonik,
"Aurora: A new model and architecture for data stream
consumption? management", The VLDB Journal, vol. 12(2), pp.120-139,
B) Compared with the traditional network seven-layer 2003.
architecture, what should be the rationalized and efficient [8] R. Motwani, J. Widom, A. Arasu, B. Babcock, S. Babu, M.
layering in WSNs? Datar, G. Manku, C. Olston, J. Rosenstein, and R. Varma,
C) Most researchers are focusing on the energy "Query processing, resource management, and
efficiency protocols. It is observed that most results are approximation in a data stream management system", The P1
Biennial Conference on Innovative Data Systems Research,
empirical with very little analytical work done. Improved or Asilomar, USA, pp.245-256, 2003.
new protocols should be developed to address higher level [9] S. Chandrasekaran, 0. Cooper, A. Deshpande, M.J. Franklin,
of robustness, scalability which involves more theoretical J.M. Hellerstein, W. Hong, S. Krishnamurthy, S. R. Madden,
works based on computational geometry, graph theory and V. Raman, F. Reiss, and M.A. Shah, "TelegraphCQ:
Continuous dataflow processing for an uncertain world", The

212
jS' Biennial Conference on Innovative Data Systems [24] F. Ye, H. Luo, J. Cheng, S. Lu, and L. Zhang, "A Two-Tier
Research, Asilomar, USA, pp.269-280, 2003. Data Dissemination Model for Large-scale Wireless Sensor
[10] S. Hedetniemi and A. Liestman, "A survey of gossiping and Networks", The 8th ACM/IEEE International Conference on
broadcasting in communication networks", Networks, vol. Mobile Computing and Networking, Atlanta, USA, pp.l148-
18(4), pp.3 19-349, 1988. 159, 2002.
[11] 1. Foster, C. Kesselman, S. Tuecke, "The anatomy of the [25] B. Karp and H.T. Kung, "GPSR: Greedy Perimeter Stateless
Grid: enabling scalable virtual organizations", International Routing for wireless networks", The 6h ACM/IEEE
Journal of Supercomputer Applications, vol. 15(3), pp.200- International Conference on Mobile Computing and
222, 2001. Networking, Boston, USA, pp.243-254, 2000.
[12] D. Abadi, Y. Ahmad, M. Balazinska, U. CJetintemel, M. [26] Y. Xu, J. Heidemann, and D. Estrin, "Geography-informed
Cherniack, J.-H. Hwang, W. Lindner, A. Maskey, A. Rasin, energy conservation for ad hoc routing", The 7th ACM/IEEE
E. Ryvkina, N. Tatbul, Y. Xing, and Stan Zdonik, "The International Conference on Mobile Computing and
Design of the Borealis Stream Processing Engine", The 2nd Networking, Rome, Italy, pp.70-84, 2001.
Biennial Conference on Innovative Data Systems Research, [27] A. Boukerchey, I. Chatzigiannakis and S. Nikoletseas,
Asilomar, USA, pp.277-289, 2005. "Efficient data propagation protocols in wireless sensor
[13] A. Demers, J. Gehrke, R. Rajaraman, N. Trigoni, and Y. networks", Wireless communications systems and networks,
Yao, "The Cougar Project: A Work-In-Progress Report", ISBN: 0-306-48190-1, pp.23-51, Plenum Press, New York,
SIGMOD Record, vol. 34(4), 2003. USA, 2004.
[14] P. Bonnet, J. Gehrke, T. Mayrand and P. Seshadri, "Query [28] K. Sohrabi, J. Gao, V. Ailawadhi and G. J. Pottie, "Protocols
Processing in a Device Database System", Technical Report, for Self-Organization of a Wireless Sensor Network", IEEE
TR99-1775, Cornell University, 1999. Personal Communications, vol. 7(5), pp.16-27, 2000.
[15] A. Deshpande, C. Guestrin, S. R. Madden, J. M. Hellerstein [29] E. Jovanov, "Patient monitoring using personal area
and W. Hong "Model-Driven Data Acquisition in Sensor networks of wireless intelligent sensors", Biomedical
Network", The 30h VLDB Conference, Toronto, Canada, pp. Sciences Instrumentation, vol. 37, pp.373-378, 2001
588-599, 2004. [30] A. Kevin, Delin and S. P. Jackson "Sensor Web for In Situ
[16] S. Madden, M. Franklin, J. Hellerstein, W. Hong, "The Exploration of Gaseous Biosignatures", IEEE Aerospace
design of an acquisitional query processor for sensor Conference, Big Sky, USA, pp.465472, 2000
networks", The ACM SIGMOD of International Conference [31] K.A. Delin, S.P. Jackson, and R.R. Some, "Sensor Webs",
on Management ofData, San Diego,USA, pp.491-502, 2003. NASA Tech Briefs, vol. 23, 1999, 80.
[17] S. D. Viglas J. F. Naughton, "Rate-based query optimization [32] C.C Shen, C. Srisathapomphat and C. Jaikaeo, "Sensor
for streaming informnation sources", The ACM SIGMOD of Information Networking Architecture and Applications",
International Conference on Management of Data, Madison, IEEE Personal Communications, vol. 8(4), pp.52-59, 2001.
USA, pp.37-48, 2002. [33] J. Elson, L. Girod, D. Estrin, "EmStar: Development with
[18] W. Heinzelman, J. Kulik, and H. Balakrishnan, "Adaptive High System Visibility", IEEE Wireless Communications,
Protocols for Information Dissemination in Wireless Sensor vol. 11(6), pp.70-77, 2004.
Networks", The 5th ACM/IEEE International Conference on [34] P. Buonadonna, D. Gay, J. M. Hellerstein, W. Hong and S.
Mobile Computing and Networking, Seattle, USA, pp.1 74- Madden, "Task: Sensor Network in a Box", 2`d European
185, 1999. Workshop on Wireless Sensor Networks, Istanbul, Turkey,
[19] C. Intanagonwiwat, R. Govindan, and D. Estrin, "Directed 2005.
Diffusion: A Scalable and Robust Communication Paradigm [35] P. Pietzuch, J. Shneidman, M. Roussopoulos, M. Seltzer and
for Sensor Networks", The 6th ACM/IEEE International M. Welsh, "Path Optimization in Stream-Based Overlay
Conference on Mobile Computing and Networking, Boston, Networks", Harvard Technical Report TR-26-04, 2004.
USA, pp.55-67, 2000. [36] R. Huebsch, B. Chun, J. M. Hellerstein, B. T. Loo, P.
[20] W. Heinzelman, A. Chandrakasan, and H. Balakrishnan, Maniatis, T. Roscoe, S. Shenker, I. Stoica and A. R.
"Energy-Efficient Communication Protocol for Wireless Yumerefendi, "The Architecture of PIER: an Internet-Scale
Microsensor Networks", The 33rd Annual Hawaii Query Processor", 2nd Biennial Conference on Innovative
International Conference on System Sciences, Maui, USA, Data System Research, Asilomar, USA, pp28-43. 2005.
pp.3005-3014, 2000. [37] I. Foster, J. Frey and S. Graham et al, "Modeling Stateful
[21] A. Manjeshwar and D.P. Agrawal, "TEEN: A routing Resources with Web Services", White Paper, 2004.
protocol for enhanced efficiency in wireless sensor [38] J. Shneidman, P. Pietzuch, J. Ledlie, M. Roussopoulos, M.
networks", The 15'h International Parallel and Distributed Seltzer and M. Welsh, "Hourglass: An Infrastructure for
Processing Symposium Workshops, San Francisco, USA, Connecting Sensor Networks and Applications", Harvard
p.30189a, 2001. Technical Report, TR-21-04, 2004.
[22] A. Manjeshwar, Q. A. Zeng, D. P. Agrawal, "An analytical
model for information retrieval in wireless sensor networks
using enhanced APTEEN protocol", IEEE Transactions on
Parallel and Distributed Systems, vol. 13(12), pp.1290-1302,
2002.
[23] S. Lindsey and C. Raghavendra, "Pegasis: Power efficient
gathering in sensor information systems", IEEE Aerospace
Conference, Big Sky, USA, pp.31 120-31130, 2002.

213

Potrebbero piacerti anche