|
       
| |
Workshop Program:
8:30 AM - 8:45 AM
Opening Remarks/Introduction
Qian Wang
(The Penn State University, USA) and Sharad Singhal (HP Labs, USA) (Workshop
Chairs)
8:45 AM - 9:45 AM: Keynote Speech
Cyber-physical Systems Research
Challenges
Dr. Jeannette M. Wing, Assistant Director, CISE, NSF,
and President's Professor of Computer Science, CMU
9:45 AM - 10:00 AM
Morning Break
10:00 AM - 12:00 PM
Session 1: Control Methodologies and Tools
Chair: Yann Labit (Laas-cnrs,
France)
- 10:00 AM
Simplified Control for Complex IT Systems Using Dimension Reduction
- Xue Liu (Mcgill University, Canada); Xiaoyun Zhu
(Hewlett Packard Laboratoties, USA); Jianguo Yao (McGill University,
Canada); Zhikui Wang (Hewlett-Packard Laboratories, USA); Sharad Singhal
(HP Labs, USA)
Automated management of complex information
technology (IT) applications and systems require dynamic
configuration of both application-level and system-level parameters.
The existence of large number of tunable parameters makes it
difficult to design a feedback controller that adjusts these
parameters effectively in order to achieve application-level quality
of service (QoS) targets. In this paper, we introduce a new approach
for simplified control of complex IT systems based on dimension
reduction techniques. It combines online selection of critical
control knobs through Lasso — a powerful L1-constrained fitting
methods, and adaptive control of the identified knobs. The latter
relies on the online estimation of the input-output model with the
selected control knobs using the recursive least square (RLS) method
and a self-tuning linear quadratic (LQ) optimal controller for
output regulation. The results of a simulation study in Matlab are
presented to demonstrate the effectiveness of our approach.
 PDF version of
paper
- 10:30 AM A model
integrated framework for designing self-managing computing systems
- Jia Bai (Vanderbilt University, USA); Sherif
Abdelwahed (Mississippi State University, USA)
Control-theoretic concepts have recently been
investigated and applied successfully to automate the management of
computation systems. However, most domain engineers are not familiar
with models, theories, and algorithms of this promising technology.
To facilitate the adoption of control-based technology for
self-management in computation systems, this paper presents a model
integrated framework, referred to as the Automatic Control Modeling
Environment (ACME). ACME is a domain-specific graphical modeling
environment encapsulated with automated synthesis tools. It allows
domain engineers to develop models that capture important behavioral
aspects of general computation systems as well as their performance
requirements and operational constraints. The synthesis component
automatically generate the code for an appropriate controller based
on a given system model and specifications. A case study of an
online power management is used to demonstrate the application of
ACME.PDF version of
paper
- 11:00 AM Frequency-Domain
Reliability Analysis and Modeling of Networked Control Systems
- Long Zheng (McGill University, Canada); Xue Liu (Mcgill
University, Canada)
Recently, advances of control theories and computer
networking technologies have made it attractive to construct complex
networked control systems for many applications. These systems
consist of a multitude of Commercial-Off-the-Shelf (COTS) hardware
and network components that may fail eventually. As networks of
computer control systems are being employed in more and more areas
where a single malfunction could lead to catastrophic consequences,
interests have grown in conducting efficient reliability analysis of
these systems. In this paper, we discuss the reliability analysis
and modeling of Networked Control Systems (NCSs) with the serial,
parallel, feedback, and back-up structures. A novel frequency-domain
block diagram (FBD) model for network reliability analysis is
presented. Using the mutual transformations of reliability functions
between the time-domain and the frequency-domain in our model, the
reliability of networked control system is analyzed. Numerical
experiment results illustrate the feasibility and effectiveness of
this approach.paper
not available
- 11:30 AM Towards
Fault-Adaptive Control of Enterprise Computing Systems---A Position
Paper
- Dara Kusic (Drexel University, USA); Nagarajan
Kandasamy (Drexel University, USA); Sherif Abdelwahed (Mississippi State
University, USA); Geoff Jiang (NEC Labs America, USA)
There is a growing interest in implementing online
control frameworks that manage distributed computing systems for
power and performance objectives. While such frameworks continuously
manage the system to optimize resource allocation and respond to
dynamic environment input, they often rely upon static models of
application behavior that do not adapt to slow behavior changes that
occur during normal operation. By introducing adaptive models that
dynamically adjust to the changing performance profile of an
application, a robust controller can maintain performance objectives
through normal changes that can occur in production and those
introduced by software errors. In this paper, we characterize the
effects of events that change an application's performance profile
over time. Such studies motivate the need for model-adaptive control
to maintain system power and performance objectives over time under
dynamic operating conditions.
PDF version of
paper
12:00 PM - 1:00 PM
Lunch Break
1:00 PM - 2:30 PM
Session 2: Resource and Power Management
Chair: Sherif Abdelwahed
(Mississippi State University, USA)
- 1:00 PM Optimizing
Concurrency Levels in the .NET ThreadPool: A Case Study of Controller
Design and Implementation
- Joseph Hellerstein (Microsoft, USA); Vance
Morrison (Microsoft, Venezuela); Eric Eilebrecht (Microsoft, USA)
This paper presents a case study of developing a hill
climbing concurrency controller for the .NET ThreadPool. The intent
of the case study is to provide insight into software considerations
for controller design, testing, and implementation. The case study
is structured as a series of issues encountered and approaches taken
to their resolution. Examples of issues and approaches include: (a)
addressing the need to combine a hill climbing control law with
rule-based techniques by the use of hybrid control; (b) increasing
the efficiency and reducing the variability of the test environment
by using resource emulation; and (c) effectively assessing design
choices by using test scenarios for which the optimal concurrency
level can be computed analytically and hence desired test results
are known a priori. We believe that these issues and approaches have
broad application to controllers for resource management of software
systems.PDF
version of paper
- 1:30 PM Applying Kalman
Filters to Dynamic Resource Provisioning of Virtualized Server
Applications
- Evangelia Kalyvianaki (University of Cambridge,
United Kingdom); Themistoklis Charalambous (University of Cambridge,
United Kingdom); Steven Hand (University of Cambridge, United Kingdom)
Resource management in virtualized data centres is
important and challenging, particularly when dealing with complex
multi-tier server applications and fluctuating workloads. In this
paper, we use control theory to build two controllers based on
Kalman filters which monitor and vary CPU allocations across
application tiers. Our approach (a) tracks utilisation patterns over
noisy data, (b) considers the resource coupling among tiers and
collectively allocates resources to them, and (c) adapts to workload
conditions through an on-line parameter estimation mechanism. An
initial experimental evaluation on a multi-tier server application
shows that our controllers work effectively.
PDF version of
paper
- 2:00 PM Feedback Control
Algorithms for Power Management of Servers
- Zhikui Wang (Hewlett-Packard Laboratories, USA);
Xiaoyun Zhu (Hewlett Packard Laboratoties, USA); Cliff McCarthy (Hewlett
Packard, USA); Partha Ranganathan (Hewlett Packard Labs, USA); Vanish
Talwar (HP Labs, USA)
Power delivery, electricity consumption and heat
management are becoming key challenges in data center environments.
To address these challenges, solutions have been developed for
average and peak power management in the data center. However, these
individual solutions are not coordinated resulting in interference
and inefficiency. In this paper, we focus on feedback control
algorithms for unified power management of a group of servers
through frequency scaling knobs. We present individual efficiency
and server capping algorithms, as well as their combined deployment
through a unified control architecture. We study the dynamic control
algorithms with qualitative and quantitative analysis. The overall
results through trace-driven simulations show that the servers under
integrated control algorithms achieve good tradeoff among power
capping, efficiency and application performance.
Paper
not available
2:30 PM - 2:45 PM
Afternoon Break 1
2:45 PM - 4:15 PM
Session 3: Control of Sensors and Networks
Chair: Zhikui Wang (Hewlett
Packard Laboratories, USA)
- 2:45 PM Estimating
Sensor Lifetime using an Event Based Control Strategy
- Agostino Capponi (Caltech, USA); Ling Shi
(California Institute of Technology, USA); Concetta Pilotto (California
Institute of Technology, USA); Richard Murray (California Institute of
Technology, USA)
We study the problem of estimating the sensor
lifetime following an event based control strategy. The sensor is
used to observe a continuous process and the measurement produced by
the sensor is sent to a base station only if the observed parameter
deviates from the model by a given threshold. Using this event based
control strategy, we are able to show that the expect lifetime of
the sensor grows quadratically with the size of the monitoring
region, and is inversely proportional to the square of the diffusion
coefficients of the process and measurement noises. Simulations are
provided to verify the theory developed.
PDF version of
paper
- 3:15 PM Sensor Network
Lifetime Maximization Via Sensor Trees Construction and Scheduling
- Ling Shi
(California Institute of Technology, USA); Agostino Capponi (Caltech,
USA); Karl Henrik Johansson Karl Henrik (Royal Institute of Technology,
Sweden); Richard Murray (California Institute of Technology, USA)
In this paper we consider state estimation carried
over a sensor network. A fusion center forms a local multi-hop tree
of sensors and fuses the data into a state estimate. A set of sensor
trees with desired properties is constructed, and those sensor trees
are scheduled in such a way that the network lifetime is maximized.
The sensor tree construction and scheduling algorithms are shown to
have low polynomial time complexity which lead to efficient
implementation in practice. The scheduling algorithm is also shown
to return the optimal solution. Examples are provided to demonstrate
the algorithms.PDF version of
paper
- 3:45 PM Network Anomaly
Estimation for TCP/AQM Networks using an Observer
- Yann Labit (Laas-cnrs, France); Frederic
Gouaisbaut (Laas-cnrs, France); Yassine Ariba (Laas-cnrs, France)
Network anomaly detection is an active research area
in network community. Researchers have approached this problem using
various techniques such as artificial intelligence, machine
learning, state machine modeling, statistical approaches. The
purpose of this preliminary work is to design an observer for
network anomaly estimation for TCP/AQM (Transmission Control
Protocol/Active Queue Management) networks using time delay system
approach. Collaborating an observer with an AQM, one can detect and
estimate anomalies considered as perturbation for the network. We
illustrate the effectiveness of results via SIMULINK and the NS-2
simulator.PDF version of
paper
4:15 PM - 4:30 PM
Afternoon Break 2
4:30 PM - 5:15 PM
Panel Discussion
Real-world considerations for Using
Formal Techniques for Feedback Control in Computing and Network Systems
Panel Chair: Joeseph Hellerstein, Microsoft
5:15 PM - 5:30 PM
Closing Remarks
Qian Wang
(The Penn State University, USA) and Sharad Singhal (HP Labs, USA) (Workshop
Chairs)
| |
|
