|
|
|||||||||||||||||||||||||||||||||||||||||||
|
Temperature-Aware
Task Scheduling for Multiprocessor SoCs In
deep submicron circuits, thermal hot spots and temperature variations have brought
new challenges in reliability, performance, cooling costs and leakage power.
Conventional thermal management sacri_ces performance to control the thermal
behavior by slowing down or stalling the processors when a critical
temperature threshold is exceeded. Moreover, such techniques do not target
minimizing the temporal and spatial variations in temperature, which impact
system reliability adversely. In this project, we explore
temperature-aware task scheduling for multiprocessor systems-on-a-chip (MPSoC).
We first establishe a baseline for dynamic scheduling techniques by solving
the task scheduling problem using integer linear programming (ILP). The ILP
solution is guaranteed to be optimal given execution times, deadlines and
temperature estimates per task. Next, we design and evaluate OS-level dynamic
scheduling policies with negligible performance overhead. We show that, using
simple-to-implement scheduling policies that make decisions based on
temperature measurements, frequency of high-magnitude thermal cycles and
spatial gradients can be decreased by around 50% and 70% respectively, in
comparison to state-of-the-art schedulers. We also enhance reactive thermal
management strategies such as dynamic thread migration with our scheduling
policies. This way, hot spots and temperature variations are decreased, and
the performance cost is significantly reduced. Resource Management in
Wireless Sensor Networks
The goal of this
project is to design scheduling and routing algorithms capable of supporting
quality of service (QoS) requirements for users of HPWREN. HPWREN today supports a large
assortment of sensor applications with varying resource needs, such as large
bandwidth requirements of Palomar observatory, low bandwidth but tight
real-time traffic deadlines of seismic sensor nodes, and long battery
lifetime essential to small and remotely deployed weather stations. Some of
the applications have stringent QoS requirements. A good example is providing
timely information to the emergency response personnel on spread of a big
forest fire or sending an alarm at the start of tremors that might lead to a
large earthquake. Although in these situations the individual sensor readings
do not consume much bandwidth (e.g. seismic sensor produce maximum 10kbps),
the timely delivery of the data, in the midst of other data already present
in the network, is of critical importance. The results of our work will
ensure that high priority sensor traffic has all the resources needed to get
to the intended destination without neglecting lower priority data. Wireless
sensing with energy harvesting for structural health monitoring: SHiMmer In this
project we study the feasibility of using energy harvesting to power sensor
nodes that are capable of sophisticated computation needed in structural
health monitoring applications. To completely avoid the use of batteries, all
power harvested by the solar cells is stored in supercapacitors. In addition,
to further maximize the energy available for computation, the node has near
zero stand-by power consumption. When active, it performs analysis of the
health condition of the structure via active sensing techniques based on Lamb
waves. These techniques are implemented by piezoelectric patches (PZTs)
organized in a grid and placed on a portion of the structure. PZT are devices
which show a coupling between mechanical and electrical domains: they produce
an electrical signal if they are bent and they vibrate if excited by a proper
wave. By analyzing the response of a sensor to the vibration induced in the
structure by the actuator and selecting specific PZTs for actuating and
sensing, it is possible to determine the location and recognize the kind of
the damage. Reconfigurable
Computing in Embedded Systems for Healthcare Applications Easy run-time reconfigurability
is a very desirable quality in many wireless sensor applications. In our
project we show how a reconfigurable embedded sensor node, SunSPOT, can
significantly extend its set of capabilities through run-time
reconfiguration. We illustrate this using a portable electrocardiogram
application whose job is to track the EKG patterns in patients and detect any
potential problems. Since a set of possible anomalies is much larger than the
size of available storage on the node, our design starts out with a subset of
algorithms capable of detecting the most likely problems. As the EKG is being
analyzed, the device can then reconfigure itself as necessary to be able to
detect all different types of heart conditions at very low cost. |
||||||||||||||||||||||||||||||||||||||||||
|
|
|
||||||||||||||||||||||||||||||||||||||||||
