Spring Banquet
Hi everyone. This forum will contain information on the Spring Banquet. The banquet is scheduled for Friday, April 4th, from 6pm - 9pm. I hope to see you guys there!
For information on Dr. Cheng, visit this website.
Power-Aware Scheduling and Quality-of-Service Optimization:
Two Enabling Technologies for Next-Generation Embedded Real-Time Systems
By Dr. Albert Cheng
Abstract
Reducing power consumption has been recognized as an important role
in the design of portable embedded and real-time systems.
Many power-saving techniques based on Dynamic Voltage Scaling (DVS)
have been proposed to reduce the dynamic power dissipation
due to the switching activities in the CMOS circuit of the embedded
microprocessor. DVS saves energy
by slowing down the processor into multiple frequency or voltage levels.
However, the worst-case execution times (WCETs) of the application tasks
scheduled by DVS typically must be
known ahead of time to ensure their schedulability.
However, a system's workload may change significantly in practice and
the real execution times (RETs) of the tasks may significantly
differ from their corresponding WCETs.
The first part of this talk presents a novel Dynamic-Mode EDF
(earliest-deadline-first) scheduling algorithm for dynamic workloads.
This algorithm estimates the RETs of application tasks
and exploits the actual slack at run-time.
The challenge is that only one combination of the number of
modes/speeds, speed-switching
transition points, and the frequency scaling factor for each mode
can lead to the best energy saving.
Experimental results show that
our Dynamic-Mode DVS algorithm achieves significant energy savings
over the traditional two-mode DVS scheme on hard real-time systems.
In the second part of this talk we turn our attention to
Quality-of-Service (QoS) optimization.
In many real-time and embedded applications, it is desirable
to achieve the highest QoS possible given the
particular system configuration and to maintain this QoS over time
despite intervals of system overload.
Increasing the QoS later following a low-QoS interval may not be
acceptable.
However, many current system applications have not fully addressed
these issues.
The objective is to optimize the QoS and to maintain
its stability while guaranteeing a schedulable task set in
these firm real-time systems.
We introduce techniques to handle overloads,
especially important in real-time systems since they must continue
to function during high-demand or extreme conditions.
This part focuses on the (m,k)-firm scheduling model and its
extensions.
In this model, m out of k jobs must meet their deadlines
in any interval of length k.
Taste of Italiano Buffet Menu
Caesar Salad
Italian vegetables
Rotelli pasta
Meat sauce & Alfredo sauce
Vegetarian lasagna
Parmesan bread sticks with butter
Italian Creme Cake