Aly, Mohamed Abdel Mohsen
(2009)
Load Balancing Hotspots in Sensor Storage Systems.
Doctoral Dissertation, University of Pittsburgh.
(Unpublished)
Abstract
Sensor networks provide us with the means of effectively monitoring and interacting with the physical world. A sensor network usually consists of a large number of small inexpensive battery-operated sensors deployed in a geographic area. This dissertation considers a sensor network deployed to monitor a disaster area. First responders continuously issue ad-hoc queries while moving in the disaster area. In such an environment, it is often more beneficial to store sensor readings and process ad-hoc queries within rather than outside the sensor network.Recently, this led to an increased popularity of Data-Centric Storage (DCS).A DCS scheme is based on a mapping function from readings to sensors based on the attribute values of each reading. This mapping function defines the DCS index structure.Two significant problems arising in this DCS network model due to data and traffic skewness are storage hotspots and query hotspots. Storage hotspots are formed when many sensor readings are mapped for storage to a relatively small number of sensor nodes. Query hotspots occur when many user queries target few sensor nodes. Both types of hotspots are hard to predict. Storage hotspots result in an uncontrolled reading shedding that decreases the Quality of Data (QoD). Due to the limited wireless bandwidth of sensors, hotspots decrease QoD by increasing collisions (thus losses) of reading/query packets. When lasting long enough, hotspots affect the Quality of Service (QoS) by unevenly depleting energy in the sensor network.This dissertation addresses both problems of hotspots through load balancing. The main dissertation hypothesis is that data migration resulting from local or global load balancing of the DCS index structure can effectively solve the hotspot problems. The contributions of this dissertation lie in developing two schemes, namely, the Zone Sharing/Zone Partitioning/Zone Partial Replication (ZS/ZP/ZPR) scheme and the K-D tree based Data-Centric Storage (KDDCS) scheme. ZS/ZP/ZPR detects and decomposes both types of hotspots through load balancing in the hotspot area. KDDCS avoids the formation of hotspots through globally load-balancing the underlying DCS index structure. Experimental evaluation shows the effectiveness of the proposed schemes in coping with hotspots in comparison to the state-of-the-art DCS schemes.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
5 June 2009 |
| Date Type: |
Completion |
| Defense Date: |
2 December 2008 |
| Approval Date: |
5 June 2009 |
| Submission Date: |
30 November 2008 |
| Access Restriction: |
No restriction; Release the ETD for access worldwide immediately. |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
Dietrich School of Arts and Sciences > Computer Science |
| Degree: |
PhD - Doctor of Philosophy |
| Thesis Type: |
Doctoral Dissertation |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
Data-Centric Storage; Load Balancing; Query Hotspot; Sensor Networks; Distributed Algorithms; Storage Hotspots |
| Other ID: |
http://etd.library.pitt.edu/ETD/available/etd-11302008-174214/, etd-11302008-174214 |
| Date Deposited: |
10 Nov 2011 20:06 |
| Last Modified: |
15 Nov 2016 13:52 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/9882 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](/images/feed-icon-32x32.png){kind=icon}