Network-Aware Content Shaping,
Energy Efficient Data Services and energy consumption
modeling for Wireless Networks
In this project, we proposed
to study techniques to transform and deliver diverse
content adaptively so as to minimize the energy
consumed by the wireless handheld. Besides the
content transformation techniques themselves,
we had to develop fast and accurate handheld energy
modeling techniques that can be used to drive
the content transformation algorithms.
In terms of energy modeling,
we extended the energy model developed earlier
for Hyper-Text Transport Protocol (HTTP)-based
textual data communication, to include multimedia
access,
specifically image access, and to consider dynamic
channel variations. In addition to the energy
spent in communicating the data, processing multimedia
data at the handheld can constitute a significant
portion of the total energy consumption. Hence,
the computational energy model is extended to
include the energy consumed in decompressing images.
Next, we considered the effect of other image
compression specific parameters as the inputs
to the energy model in addition to the size of
the objects accessed, as specified in our earlier
model. Additionally, channel condition variations
can affect the energy consumption significantly.
For instance, when the channel condition, represented
by the Signal-to-Noise Ratio (SNR) degrades, the
energy
consumed in accessing the multimedia object increases.
This is due the fact that the transmission power
used and the number of retransmissions required
to communicate increases under poor
channel conditions.
In order to consider the above
enhancements, we measured energy consumption under
different controlled conditions using our data
acquisition platform. Based on the measured data,
we performed regression analysis to develop the
energy model considering different input variables
identified before. The input parameters that we
considered are the service specific parameters
(volume of data requested, image size, image compression
parameter) and networkrelated parameters (SNR).
The resulting energy model is validated and used
with adaptive image shaping techniques developed
in a separate project. The use of the energy model
in guiding the
adaptive image shaping technique leads to significant
savings in energy consumption with minimal degradation
to image quality. For example, for a medical image
of size 160x160, our adaptive image delivery techniques
consume 6.4J of energy compared to 11.6J without
adaptation in Palm.Net access network, while degrading
image quality from 40.3dB to 32.1dB. After studying
the effect of network conditions on energy consumption,
we have started to investigate the effect of another
network condition, i.e., error patterns on application
data. Below, we briefly describe the goal of the
current efforts and the architecture proposed.
One significant bottleneck in
enabling high quality multimedia applications
is the dynamic error condition caused by wireless
channel variation. In order to address the erroneous
channel conditions, several link and physical
layer error control techniques have been proposed
to counter the presence of errors by introducing
redundancies in transmission. However, the low
layer techniques have high associated communication
costs in terms of energy and latency due to the
added data transmission. Additionally, these physical/link
level techniques are oblivious to application
requirements such as inherent error resiliency
in application data.
We believe that the overhead
of error control can be reduced significantly
by adapting the error control mechanisms based
on current context represented by application
properties and wireless
channel conditions. We are developing a Context-aware
Error Control (ConECt) framework that uses application
level information to enable low-cost error control
through proper selection and configuration of
error control mechanisms. ConECt would enable
trade-off between an
application’s Quality of Content (QoC) and
communication cost in diverse wireless contexts.
Our approach to achieve the
above goal consists of two functional steps: (1)
characterization of the effect of different error
profiles and error control mechanisms on application’s
data in terms of Quality of Content (QoC) and
communication cost (energy/latency/access cost),
and (2) dynamic selection of the error control
mechanism at runtime using the pre-characterized
application’s error-effect models developed
in the first step. The proposed ConECt framework
uses current channel conditions to select appropriate
error control mechanism in order to reduce communication
cost without affecting QoC significantly. The
framework is compatible with current wireless
standards as it chooses an error control mechanism
from the set of choices allowable by the standard
specifications. For example, the majority of 3G
wireless data standards currently have support
for multiple channel coding algorithms (i.e. convolutional
or turbo), and different coding rates at the physical
layer. As the ConECt framework does not require
any modification to the existing functionality,
it can be deployed easily on current wireless
data networks. We are currently evaluating the
above framework under diverse channel conditions
and data types.
