Date of Award

5-2012

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Legacy Department

Computer Science

Advisor

Hallstrom, Jason O

Committee Member

Srimani , Pradip K

Committee Member

Goddard , Wayne D

Committee Member

Malloy , Brian A

Abstract

Despite the increasing application of embedded wireless systems, developers face numerous challenges during the design phase of the application life cycle. One of the critical challenges is ensuring performance reliability with respect to radio link quality. Specifically, embedded links experience exaggerated link quality variation, which results in undesirable wireless performance characteristics. Unfortunately, the resulting post-deployment behaviors often necessitate network redeployment.
Another challenge is recovering from faults that commonly occur in embedded wireless systems, including node failure and state corruption. Self-stabilizing algorithms can provide recovery in the presence of such faults. These algorithms guarantee the eventual satisfaction of a given state legitimacy predicate regardless of the initial state of the network. Their practical behavior is often different from theoretical analyses. Unfortunately, there is little tool support for facilitating the experimental analysis of self-stabilizing systems.
We present two contributions to support the design phase of embedded wireless system development. First, we provide two empirical models that predict radio-link quality within specific deployment environments. These models predict link performance as a function of inter-node distance and radio power level. The models are culled from extensive experimentation in open grass field and dense forest environments using all radio power levels and covering up to the maximum distances reachable by the radio.
Second, we provide a simulation framework for simulating self-stabilizing algorithms. The framework provides three feature extensions: (i) fault injection to study algorithm behavior under various fault scenarios, (ii) automated detection of non-stabilizing behavior; and (iii) integration of the link quality models described above. Our contributions aim at avoiding problems that could result in the need for network redeployment.

Share

COinS