Transcript Document

DFG/GENI
Doctoral Consortium,
San Juan, PR
March 13th-15th, 2011
GridStat on GENI: Simulating a Smart Power
Grid Infrastructure over GENI
Divya Giri, Ruma Paul, Haiqin Liu, Victor Valgenti, Carl Hauser and Min Sik Kim
Washington State University
Abstract
Developments in power grid measurement and
monitoring technology have enabled precise
and frequent measurement of the state of the
power grid. Modern power grid control
infrastructure are insufficient to the effective
forwarding of this information to the necessary
control facilities. The GridStat framework offers
an efficient, low-latency data forwarding
framework that can provide the necessary
Quality of Service for control facilities to
maintain sub-second status of monitored power
grid substations. However, the current GridStat
prototypes have not been tested outside local
clusters. The GENI infrastructure, in particular
PlanetLab, provide the platform through which it
is possible to test GridStat at scale and identify
problems with the current framework.
Furture Work
Synchrophasor Measurement Access to
GENI: WSU has one synchroPhasor
Measurement Unit (PMU), donated by
[5]. Our goal is to provide an interface
such that meaningful measurements
are published to GENI and can be
consumed by prospective users of this
data. Optimally, other PMUs could be
connected to GENI to create a virtual
power grid.
Improving GridStat: In addition to the
above experiments, we hope to use
the GENI infrastructure and this
implementation of GridStat in future
enhancements and evaluations of the
GridStat framework.
F
E
F
E
• Create a functioning GridStat overlay on top
of GENI
• Evaluate GridStat in face of pandemic failure
and security threats
• Develop and evaluate methods to identify and
counter failures and threats
• Develop efficient dynamic routing to meet
GridStat QoS over GENI
• Identify new instrumentation for maintaining
system awareness of GridStat
F
E
F
E
F
E
F
E
F
E
[5]
GridStat Network
Publisher
Subscriber
Substations, as well as other locales, monitor the power network.
They take measurements describing the current state of the power
grid at that location. These measurements are published to the
GridStat Network which forwards them to subscribers who gather
those measurements for an accurate view of the current state of the
power grid.
Experiments
GridStat on GENI: Recreate the work as initially
demonstrated in [1,2]. This required building
forwarding engines to run on PlanetLab
nodes. However, since PlanetLab does not
necessarily maintain required latency bounds
we added dynamic routing to detect and
change routes when latency boundaries are
exceeded.
GridStat Survivability: These tests examine the
impact of failing and compromised nodes on
the GridStat network. It consists of three
distinct batteries: Blackout attack, where n
Forwarding Engines (FE) simply cease
functioning, Starvation attack, where n FE
suffer DOS, and the Confusion attack where
n FE start propagating spurious data.
Instrumenting and Security for GridStat on
GENI: These tests will extend the work from
[3,4] to detect security threats, failing nodes,
and congested links to improve the
sustainability of GridStat. This is a repeat of
the GridStat Survivability tests with a focus of
identifying problems and routing around
them.
Research Objectives
Overview of GridStat [1,2]
P
F
F
2
2
F
F
1
3
P
S
F
F
1
3
F
F
4
4
Forwarding under two routes
F1,F2 and F1,F4. Route F1,F2 is
experiencing high delay causing
packets to miss QoS bounds.
Dynamic Routing Current & Future Experiments
S
Dynamically replace F1,F2 with F1,F3
which is maintaining QoS bounds.
• Instrumentation uses probe packets to periodically
determine latency on links (broadcast probe).
• Initial Results demonstrate some fluctuation in calculations
dependent on number of probes-per-second—examining
ways to minimize probes.
• Next step: Implement Dynamic routing and test impact of
route changes.
Dynamic Routing
GridStatTraffic is forwarded through the
Monitor Q
network. Each node collects data about
all packets arriving at each node. That
F
2
data is periodically transmitted to the
F
TrustGuard monitor [3,4]. GridStat traffic
P
1
is forwarded through the network.
F
DDoS
3
However, Large shifts in IP address can
A
be detected when a DDoS attack occurs
as per [3,4]. The monitor can detect
Basic DDoS Attack these attacks as indicated in the
preliminary results found to the right.
Detecting Purposeful Attacks against GridStat Nodes
Alarm series Minimal H series Volume (Bytes)
1st
5
4
x 10
2
0
0
50
100
150
200
250
300
350
200
250
300
350
300
350
Time (s)
5
0
0
50
100
150
Time (s)
2
Detected
attack
1
0
0
50
100
150
200
250
Time (s)
Preliminary Results for Detecting Attacks
Proposed Publications
Use of Glab/GENI Infrastructure
Forwarding Engines (FE) run as applications on
PlanetLab nodes. The FE forward power
measurements to subscribers. A QoS layer monitors
routes. If a route falls out of QoS bounds due to
congestion or attack, new routes are chosen. The
QoS layer uses active and passive monitoring.
The Research efforts have resulted in the following
proposed publications:
Conference / Journal Papers
Divya Giri and Min Sik Kim, “Monitoring and Securing Information in
Mission Critical Networks”, TBD.
Master Thesis
Q
P
publisher
QoS/Security
Q
Q
Q
S
subscriber
Layer
F
E
P
Divya Giri, “Dynamic Routing for Enhancing Security in Mission Critical
Networks”
F
E
F
E
F
E
F
E
Q
forwarding
engine
Q
S
Forwarding
Layer
Basic utilization of PlanetLab nodes
QoS
monitor
References
1. G. Harald, D. Bakken, C. Hauser, and A. Bose, “GridStat: A Flexible QoS
Managed Data Dissemination Framework for the Power Grid,” IEEE
Transactions on Power Delivery, Jan 2009.
2. C. Hauser, D. Bakken, and A. Bose, “A Failure to Communicate,” IEEE Power
and Energy, March/April 2005.
3. H. Liu, Y. Sun, V. Valgenti, and M. Kim, “Trustguard: A Flow-level Reputationbased DDoS Defense System,” Workshop on Personalized Networks, January
2011.
4. H. Liu, and M. Kim, Real-time Detection of Stealthy DDoS Attacks using TimeSeries Decomposition,” ICC 2010.
5. SEL 351, Schweitzer Engineering Laboratory (SEL), http://www.selinc.com.