Center for Future Information Technology CFIT

1
Center for Future Information Technology
(CFIT)
2
UC Davis Welcomes

• Intel Corporation
• Hewlett Packard
• Nokia
• Cisco
• Sprint
• Qualcomm
• PC-Doctor
• Gigamon Systems
• Siemens
• Huawei

3
A Glimpse ofCFIT - Related Projects

• University of California, Davis
• cfit.ucdavis.edu

4
CoRENetCognitive Radio Enabled Heterogeneous
Wireless Network
www.cs.ucdavis.edu/liu/research

• Wireless devices are
• smarter (sensing, signal processing, decision
  making, etc.)
• heterogeneous (WiFi, WiMax, Bluetooth, UWB,
  cellular, etc.)
• densely deployed
• overlapping in space and spectrum

• Objective to exploit the advances and
  heterogeneities in wireless technologies for
• dynamic and efficient resource utilization
• traffic engineering and QoS provisioning
• enabling new services while protecting legacy
  services
• Approach to understand and manipulate
  interaction coordination among wireless devices

For more information contact Xin Liu
(liu_at_cs.ucdavis.edu)
5
SWiMScalable Heterogeneous Wireless Meshes

• SWiM is a scalable mesh network architecture that
  consists of tens or hundreds of APs
• WiFi-based mesh covering a whole city (San
  Francisco, New York, or London)
• An enterprise network include company-wide
  video-conference broadcasting, file distribution,
  and other streaming services.

• Objective
• Ubiquitous high-speed wireless access
• QoS provisioning (especially for enterprise IT)
• Exploit heterogeneous wireless technologies

• Network Structure
• Portals wired APs or WiMAX base stations
• Wireless mesh nodes wireless APs, laptops or
  desktops
• Wireless clients laptop, cellphones, PDAs,
  senor devices, etc.

Wi-Fi Cellphones
Contact Xin Liu, Prasant Mohapatra
(liu,prasant_at_cs.ucdavis.edu)
6
Heterogeneous Wireless Networks

• Design Challenges
• Interference from adjacent nodes or links
• Network congestion and QoS provisioning
• Light-weight control and management

• Our Approach -- Heterogeneous Wireless Networks
• Resource Heterogeneous wireless access networks
  (WiFi, WiMax, UWB, Bluetooth, etc.)
• Traffic Engineering Relieve congestion and
  provide QoS provisioning through network
  heterogeneity
• Control Broadcast the control information for
  QoS or smart-routing

Contact X. Liu, P. Mohapatra, C-N. Chuah
(liu,prasant_at_cs.ucdavis.edu, chuah_at_ece.ucdavis.e
du)
7
Load Balancing and Smart-routing

• Load Balancing
• A threshold-based protocol
• A heuristic algorithm implemented in WiMAX base
  station

• Smart-Routing
• Congestion collection and broadcasting
• Local decision for new paths

Contact X. Liu, P. Mohapatra, C-N. Chuah
(liu,prasant_at_cs.ucdavis.edu, chuah_at_ece.ucdavis.e
du)
8
Measurement Study of Public Safety Spectrum Usage

• Objectives
• Observe and model spatial and temporal properties
  in spectrum activities using two measurement
  suites
• Devise thresholding methods to accurately
  determine activity of licensed users

• Experiment Setup
• Measurements taken in county-wide safety network
  in Howard County, MD (JHU/APL)
• Two measurement suites used to measure activity
  synchronously at different locations
• Samples taken at 64 MSamples/s

For more information contact Xin Liu, Zhi Ding
(liu_at_cs.ucdavis.edu, ding_at_ece.ucdavis.edu)
9
Smart Antenna Technologies Applied in Cognitive
Radio Systems
System Scheme

• Objectives
• Maximize the QoS of cognitive users (secondary
  users) while protecting the performance of
  existing legacy users (primary users) through
  smart antenna technology
• Design MAC Protocols to take advantages of smart
  antenna technologies

For more information contact Xin Liu, Zhi Ding
(liu_at_cs.ucdavis.edu, ding_at_ece.ucdavis.edu)
10
Attacks and Defenses of Opportunistic Scheduling
in Cellular Data Networks

• Project Summary
• Discovered vulnerabilities in 3G handoff and
  Proportional Fair algorithm.
• - Allows rouge cellular devices to usurp
  majority of time slots at the expense of other
  users.
• - Only one rouge device per cell of 50 users
  can use up to 90 of time slots.
• Propose defense strategies
• - Robust handoff-aware scheduler
• - Resilient to handoff induced average rate
  resets
• - Improve network performance during handoff
• - Priority queue
• - Round robin hybrid

attacker
For more information contact Xin Liu, Hao Chen
(liu,chen_at_cs.ucdavis.edu)
11
Robust Routing and Scheduling in Wireless Mesh
Networks

• Objective
• Find joint routing and scheduling that work well
  under traffic variations
• Motivations
• Existing optimal routing and scheduling assume
  accurate traffic information
• Measure and disseminate accurate traffic info is
  usually not possible in wireless networks
• Joint routing and scheduling must be robust to
  traffic variations
• Approach
• Optimize the worst-case performance for a range
  of traffic conditions
• An LP model which can be solved in polynomial time

• Larger w, larger traffic variations
• TORS always achieve best worst-case performance,
  and good average performance

• Even the oblivious routing only can still achieve
  considerably better performance than AODV routing

For more information contact Xin Liu
(liu_at_cs.ucdavis.edu)
12
Maximum Capacity in Multi-channel Multi-radio
Wireless Networks

• Objective
• Try to find the maximum spatial reuse of a
  topology given of channels and of radios
• Can be used as a guideline to deploy radios on
  nodes
• Also can be used to find the goodness of a
  topology

• Result
• Given the number of available channels, there
  exists an optimal number of radios to fully
  utilize them
• Small TX range and large TX range are both not
  desirable. Optimal performance is achieve when TX
  range is somewhere in between

For more information contact Xin Liu
(liu_at_cs.ucdavis.edu)
13
Energy Efficient Throughput Optimization in
Multi-hop Wireless Networks
Motivation Existing fairness objectives do not
work well in multi-hop networks maxmin
throughput fairness let routers transmit very
limited data, while maxmin time fairness
penalizes leaf nodes seriously.

Our solution lexicographical maxmin energy
efficiency throughput fairness
Benefits Achieving fair total throughput after
each user using up energy gives routers more
bandwidth which is also bounded Guarantee the
bandwidth of users far from sinks.
Fig.1. Maxmin energy efficiency throughput
fairness gives routers higher throughput while
not sacrificing bandwidth of leaf nodes


Application scenario 1 A wireless sensor
networks with Energy-constraint nodes which must
relay traffics hop-by-hop to sinks
Application scenario 2 A wireless mesh networks
organized in multi-hop topology laptops use
battery power
For more information contact Xin Liu
(liu_at_cs.ucdavis.edu)
14
Capacity of Wireless Networks with Relay Nodes

• Motivation
• Capacity of wireless networks does not scale
  well with increasing number of uses
• Node cooperation can increase network capacity

Problem What and how capacity of wireless
networks can be achieved with relay nodes
cooperation
Main Idea Local cooperation to improve spatial
reuse Long range multiple transmissions by relay
nodes
Fig 1. Dividing Unit area into squarelets, local
cooperation within the set of dark squarelets can
be simultaneous
Main result
can be achieved for randomly deployed
wireless networks and relay nodes


Fig 2. A Illustration of relay nodes cooperated
transmission for arbitrary source destination pair
For more information contact Xin Liu
(liu_at_cs.ucdavis.edu)
15
Impairment-Aware Provisioning in All-Optical
Networks

• PI Biswanath Mukherjee
• PhD students Smita Rai and Lei Song
• Sponsor Fujitsu Research Labs, Sunnyvale (and UC
  MICRO)
• Physical impairments occur in all-optical
  networks
• Explore optimal algorithms with fast execution to
  provision paths with good Optical Signal-to-Noise
  Ratio (OSNR)
• Investigate placement of 3R regenerators to
  minimize cost

16
Integrated Wireless-Optical Broadband Access
Network (WOBAN)
PI Biswanath Mukherjee PhD student Suman
Sarkar Sponsor Nokia
San Francisco WOBAN (SFNet)
Optical access costly Wireless access
spectrum-limited WOBAN high-capacity,
inexpensive solution
ONU/AP
ONU/AP
ONU/AP
Generic Architecture
Wildhorse WOBAN
17
Dynamic Optical Circuit-Switching (DOCS) Networks

• PIs Biswanath Mukherjee, Nick McKeown, Dan
  Blumenthal, John Bowers
• UCD PhD students Dragos Andrei and Cicek Cavdar
• Sponsor NSF (Grant No. 06-27081)
• Dynamic Bandwidth Allocation dial for capacity
  on demand.
• Bandwidth provided by dynamic circuits setup on
  the fly between two routers.
• Network can be dynamically reconfigured. Boundary
  routers at edges, and core of optical switches,
  with guaranteed bandwidth.

• Bandwidth is allocated based on application
  needs sub-wavelength granularity (grooming),
  waveband granularity (virtual concatenation)
• Flexible transmission rates for transferring a
  file with a required deadline
• Deadline-driven bandwidth allocation
• Time reconfiguration for demands with various
  deadlines improves network performance

18
Designing Next-Generation Access Networks
PI Biswanath Mukherjee PhD students Huan Song,
Amitabha Banerjee, Fred Clarke Sponsors NSF
(Grant No. NSF-04-35525 09/04 to 09/07)
ETRI, Korea (05/05 to 05/07)

• Multiple Gbps bandwidth for downstream and
  upstream
• Diverse topology, such as ring and tree
• Extended network coverage, up to 100 km reach

• Smooth upgrade from EPON to WDM-PON
• Effective resource allocation to end-users
  (data-traffic-aware)
• Constrain packet delay

19
Robust Architectures for Next-Generation
End-to-end Ethernet Networking (RANGEEN)

• PI Biswanath Mukherjee
• PhD student Marwan Batayneh
• Sponsor / Collaborators Siemens AG

• Extend Ethernet boundary from LAN to WAN.
• End-to-end (PC to server) Ethernet.
• Establish high-speed Ethernet tunnels (up to 100
  Gbps).
• Provide Ethernet tunnels with Carrier-Grade
  robustness.
• Design cost-efficient Carrier-Grade Ethernet.

20
Designing Next-Generation Robust Telecom Networks

• PIs Biswanath Mukherjee and Charles U. Martel
• PhD students Smita Rai, Lei Song, Van Nguyen,
  Ananya Das
• Sponsor NSF (Grant No. CNS-05-20190 9/05 to
  8/08)
• Connection availability important metric for
  reliability/robustness
• Service Level Agreements provide guarantees
• www.sprintworldwide.com/english/solutions/sla

• Multiple paths provide effective use of network
  bandwidth
• Notion of degraded service
• Develop efficient multi-path provisioning
  algorithms
• Develop (backup) reprovisioning algorithms

21
RoSE Robust, Secure Efficient Routing

• Towards Next-Generation Internet - adding
  measurability, accountability, security, and
  adaptability
• Large-scale Internet measurements/monitoring over
  production network
• Model interaction between multiple control loops
• e.g., overlay- vs. IP-layer routing
• Multi-layer information sharing
• New solutions for traffic engineering, fault
  resilience, graceful networkupgrade, security
  defense
• PI Chuah (chuah_at_ece.ucdavis.edu) Sponsors
  NSF, Sprint, NarusProject url
  http//www.ece.ucdavis.edu/rubinet/rose.html

22
Accurate Light-weight Traffic Measurement for
Effective Network Anomaly Detection

• Challenge Traffic sampling reduces measurement
  overhead but introduces bias in flow statistics
  gt hinders effective anomaly detection
• Quest Better measurement techniques
• Preserve statistics of different traffic
  subpopulation
• Minimize resource overhead (CPU, memory)
• PI Chuah (chuah_at_ece.ucdavis.edu) Sponsor
  Sprint Project url http//www.ece.ucdavis.edu/ru
  binet/sand.html

23
MINESTRONE Multimedia Streaming and Mobile
Services over Heterogeneous Wireless Networks

• Hybrid infrastructure P2P support
• Monitoring, mobility tracking, caching, traffic
  analysis, service creations
• Applications multimedia streaming, VoIP, gaming,
  e-Commerce
• PI Chuah Sponsors Hewlett Packard, UC Micro
• Project url http//www.ece.ucdavis.edu
  /rubinet/minestrone.html

24
Cooperative Peer-to-Peer Repair

• Leverage ad hoc (e.g., 802.11 based) wireless
  connections between 3G subscribers to repair
  broadcast losses

• Distributed Repair Challenges
• Network State Information Model
• Interference between Repair transmissions
• Solution
• Batch Repair Scheduling (BRS)
• Simultaneous Control and Repair

PI Chuah (chuah_at_ece.ucdavis.edu) Sponsors
Hewlett Packard, UC Micro Project url
http//www.ece.ucdavis.edu/rubinet/minestrone.html
25
OpCom Opportunistic Networking and Collaborative
Content

• Goal Efficient use of all communication
  opportunities
• Intermittent connectivity to infrastructure
• Peer-to-peer capability, primarily single hop
• Multiple wireless network interfaces
• Opportunistic content distribution
• Prototype RSS reader

Davis WiFi Coverage
PI Chuah Sponsors Sprint Project url
http//www.ece.ucdavis.edu/rubinet/opportunistic.h
tml
26
QuRiNetQuail Ridge Wireless Mesh
Networkhttp//spirit.cs.ucdavis.edu/quailridge
QuRiNet is a wide-area operational wireless mesh
network deployed near Napa Valley

• Objective
• Provide network access to ecological researchers
  remotely
• Testbed for wireless mesh research

• Facts
• Hilly terrain with heavy forest growth
• 2,000 acres of wilderness
• Variable weather and free from external signal
  interference

• Network Architecture
• Currently 11 nodes deployed
• Dual 802.11g radios per AP
• Multi-hop Multi-channel Network
• Runs entirely on solar power

For more information contact Prasant Mohapatra
(prasant_at_cs.ucdavis.edu)
27
Wireless Mesh Overlay (WiMO)

• Objective Design and develop a light-weight,
  common management framework to coordinate
  monitoring and resource control of wireless mesh
  networks

• Deploy monitoring framework in the mesh network
• Overlay based approach for information exchange
  between nodes
• Implement measurement based management framework
• Novel approach for measuring channel utilization
  of the network
• New link layer packet forwarding scheme
• Experimental implementation on our test bed

For details contact Prasant Mohapatra
(prasant_at_cs.ucdavis.edu) and Chen-nee Chuah
(chuah_at_ece.ucdavis.edu)
28
Video over Wireless Mesh Networks
Conflict Graph Based Interference Model

• Support QoS for video transfers in wireless mesh
  networks.
• Video is content-rich, but more stringent on QoS
  requirements.
• Interference and low bandwidth degrade video
  performance.
• Our approach
• An admission control scheme coupled with QoS
  routing.
• A cross-layer approach for both single-channel
  and multi-channel wireless mesh networks.

Connectivity Graph
Conflict Graph
Links become nodes
Conventional Scheme
Our Scheme
For more information contact Prasant Mohapatra
(prasant_at_cs.ucdavis.edu)
29
Admission Control and Scheduling in Multi-hop
WiMAX Networks

• Objective To find the subchannel and timeslot
  assignment scheme that meets the QoS requirements
  (bandwidth and deadline) of flows.
• Heuristic algorithm for a tree topology based
  WiMAX network
• Interference aware subchannel-timeslot assignment
• Compares scheduling available subchannels in a
  timeslot vs available timeslots for a subchannel
• Uses pre-emption to take away extra bandwidth to
  accommodate more flows if required.
• Extendible to across frame scheduling

For more information contact Prasant Mohapatra
(prasant_at_cs.ucdavis.edu)
30
Wireless Mesh HealthMeasurements, Monitoring and
Management

• Objectives
• Develop tools for diagnosing behavior and
  anomalies in the network
• Formulate new metrics for better understanding
  of the wireless network
• Improve on algorithms and techniques used for
  repairing wireless mesh networks

• Technical Approach
• Experimental comparison of current wireless
  metrics
• Control Data collection algorithms
• Management protocol for scalable passive and
  active monitoring
• Development of tools for management of WMNs

Internet
Internet
For more information contact Prasant Mohapatra
(prasant_at_cs.ucdavis.edu)
31
Multi-radio Multi-channel Wireless Mesh Networks

• Objective
• Dynamic channel assignment and interference-free
  link scheduling to optimize performance

• Approaches
• Given traffic demand
• data size minimize the time to satisfy all
• bandwidth maximize the minimal link satisfaction
  ratio
• Process
• Generate a max-flow graph and formulize into a LP
  model
• Solve LP to get a one-time-slot schedule which
  meets radio and channel constraints
• Pack one-time-slot schedules using link weight
  adjusting strategy

3,4,5
For more information contact Prasant Mohapatra
(prasant_at_cs.ucdavis.edu)
32
Underwater Wireless Networks Medium Access Control

• Applications
• Environmental monitoring
• Resource management
• Reconnaissance

• Distinguishing Characteristics
• Highly variable channel
• Large acoustic propagation delay
• Energy intensive transmissions
• Approach
• Divide time into scheduled and unscheduled
  portions
• Nodes indicate desire for unscheduled slots
• Advantages
• Tradeoff between TDMA and random access
• Nodes gain knowledge of unscheduled slot
  contention

D
A
C
B
E
B and E know only they contend for unscheduled
slots
For more information contact Prasant Mohapatra
(prasant_at_cs.ucdavis.edu)
33
Joint MIMO Detection and LDPC Code Design for
MIMO Wireless Systems (Z. Ding and S. Lin with
Intel 06-07)

• Joint detection for broadband services
• Tailors code design for MIMO wireless

Much better coverage in WiMAX and 802.11n WIFI
Contact Z. Ding
34
Advanced Transceiver Design Integration with
ARQ (Huaiwei 07-08)
data
Channel distortion
Joint receiver and decoder
Rate reduction
Channel distortion
2nd transmission

• Scalable, efficient ARQ to improve wireless
  throughput
• Joint detection and decoding amenable to turbo
  processing

Contact Z. Ding
35
FIREMAN Firewall Configuration Verification

• Config is error-prone
• Policy violation
• Inconsistency
• Inefficiency
• Intra Inter-firewall
• Borrow model checking techniques from software
  programming

• Modeling and Validating Firewall Rules
• Static analysis (control data flow)
• Scalable Packet space exploration
• Features full coverage, scalable, efficient

• PIs Chuah, Chen, and Su Sponsor NSFProject
  url http//www.ece.ucdavis.edu/rubinet/fireman.ht
  ml

36
Examining Vulnerabilities in Opportunistic
Scheduling In Cellular Networks
Hao Chen, Xin Liu, Radmilo Racic
University of California, Davis
hchen_at_cs.ucdavis.edu
http//zeus.cs.ucdavis.edu/cellSecurity/

• Vulnerability
• UEs (user equipment) report channel conditions
  (CQI) to base station
• Base station (BS) schedules UEs based on CQI
• BS trusts UE to report correct CQI
• But UE can lie!
• Attack
• Malicious UE usurps scheduling slots by
• Reporting fabricated CQIs
• Exploiting handoff procedure
• Simulation in a cell of 50 users
• One attacker can usurp 90 slots
• Attack consequences
• Severely degrade victim users network access
• Disrupt time-sensitive services, e.g. VoIP
• Defense
• Develop robust Proportional Fair (PF) scheduler
• Resilient to fabricated CQI

37
Detect Web Spam and Track Spammers
Hao Chen and Yuan Niu
University of California, Davis
hchen_at_cs.ucdavis.edu
http//wwwcsif.cs.ucdavis.edu/niu/re-spam.html

• Objectives
• Study modus operandi of Web forum spammers
• Detect forum spam to improve Web search quality
• Identify major spammers

Doorway Pages
Ads and money
User attention
Redirectors
Aggregators
Syndicators
3. Propagates splog URL
Advertisers
Comment Spam
Search Engine
Five-Layer Model of Forum Spam
Returns
2. Spreads splog URLs

• Approach
• Discover spammer domains
• Using context-based analysis
• Focusing on redirection cloaking
• Follow the flow of money from advertisers to
  spammers
• Propose a 5-layered model of web advertising
• Identify spammers (publishers and middlemen)

Search Results
4. Sends user to doorway URL
Doorway Pages (Splogs)
Spammer Domain
Spammer
1. Creates
5. Redirects user to
Modus Operandi of Web Forum Spammers
38
Security of Domain Name System
P

• Cache Poisoning
• DNS provides sign posts to apps
• Cache recursive referrals
• Many ways to inject poison (false records) into
  cache
• Poison stays and propagates
• DNSSEC 8yrs 0.0008 coverage

.
Proxy
inject
.com
.edu
propogate
.intel
.ucd
.nba
.mit
.cs
.ece
.www

• Antidote DNS Cross Referencing (DoX)
• DoX P2P consistency check among proxies
• Difficult for coordinated poison injection
• Check with k peers, verify records changes
• Overhead can be controlled

B
1
Check A
2
Agree
Check A
A
D
Agree
Agree
Check A
C
For more information contact Prasant Mohapatra
(prasant_at_cs.ucdavis.edu)
39
COMET Managing Geoscience Data

• COMET COast-to-Mountain Environmental Transect
• Funded through NSF Cyberinfrastructure for
  Earth-Observatories Program

COMET Coast-to-Mountain Environmental Transect
Participants Bodega Marine Lab (BML), Tahoe
Environmental Research Center (TERC), Atmospheric
Sciences, Climatology, CS_at_UCD
Objectives Develop a cyberinfrastructure that
helps studying how environmental factors and
climatic variability affect major ecosystems
along an elevation gradient from coastal
California to the summit of the Sierra Nevada.
R D data registration integration,
geospatial data, scientific data management,
workflow integration, GIS Web Services ...
Contact M. Gertz and B. Ludäscher
40
GeoStreams Satellite Data Management
Adaptive Query Processing Architectures for
Streaming Geospatial Images Funded through NSF
Information Technology Research Program

Objectives Efficient processing of continuous
queries over streaming satellite data
.
24 spectral bands 5 primary scan sectors
GOES Satellite
21Mbit/s
Receiver
GeoStreams Server
Real-time GOES Viewer
Contact M. Gertz geostreams.ucdavis.edu
Antenna (_at_Academic Surge)
Persistent storage
41
Database Security Re-engineering
Security Analysis and Re-engineering of
Databases Funded through NSF Data and
Application Security Program
Database Objects
Objectives Profile the evolution of and access
to mission critical data over time. Use data and
user profiles to adjust database security
policies and mechanisms. Focus on insider
misuse.
OS Users
Application Users
Database Users
Contact M. Gertz dbis.ucdavis.edu/research/dbsec
Access Path
Access Path Model as basis for focused data/user
profiling and reasoning
Databases Information System. M. Gertz
(GERTZ_at_UCDAVIS.EDU) B. Ludescher
(LUDAESCH_at_UCDAVIS.EDU)
42
Scientific Workflows Cyberinfrastructure
UPPER-WARE
Upperware
Upper Middleware
Middleware
Underware
NSF ITR Science Environment for Ecological
Knowledge (SEEK)
Databases Information System. M. Gertz
(GERTZ_at_UCDAVIS.EDU) B. Ludescher
(LUDAESCH_at_UCDAVIS.EDU)
43
Scientific Workflows for Phylogenetic Analyses
Aligned DNA sequences
Maximum likelihood tree (DNA)
Discrete morphological data
Maximum parsimony tree
Integrate
Consensus Tree(s)
Maximum likelihood tree (continuous characters)
Continuous characters
NSF Collaborative Research (w/ UPenn) Core
Database Technologies to Enable the Integration
of AToL Information
Actors
Datasets
Datasets
Databases Information System. M. Gertz
(GERTZ_at_UCDAVIS.EDU) B. Ludescher
(LUDAESCH_at_UCDAVIS.EDU)
44
Bioinformatics Workflow Modeling Automation
collaboration with UCD Genome Center
NSF/SEI(BIO)II A Collaborative Scientific
Workflow Environment for Accelerating
Genome-Scale Biological Research
Databases Information System. M. Gertz
(GERTZ_at_UCDAVIS.EDU) B. Ludescher
(LUDAESCH_at_UCDAVIS.EDU)
45
Kepler and Sensor Networks

• Management and Analysis of Environmental
  Observatory Data using the Kepler Scientific
  Workflow System, NCEAS, SDSC, UC Davis, OSU,
  CENS (UCLA), OPeNDAP
• standardize services for sensor networks, support
  multiple views, protocols
• COMET Coast-to-Mountain Environmental Transect,
  UC Davis, Bodega Marine Lab, Lake Tahoe Research
  Center
• study how environmental factors affect ecosystems
  along an elevation gradient from coastal
  California to the summit of the Sierra Nevada

CEOP--COMET Coast-to-Mountain Environmental
Transect
CEOP--Management and Analysis of Environmental
Observatory Data Using the Kepler Scientific
Workflow System
CEOP/REAP
Databases Information System. M. Gertz
(GERTZ_at_UCDAVIS.EDU) B. Ludescher
(LUDAESCH_at_UCDAVIS.EDU)
46
Performance Enhancement of Discrete-Event
Simulation
Norm Matloffs Current Research
http//heather.cs.ucdavis.edu/nm.html

• Many DES packages now use interpreted languages,
  e.g. Python.
• How to improve performance?
• Hybrid Use Python for most of the code, but C
  for the core operations.
• Data structure improvements too.
• Speedups of 30-60.
• Paper to appear in Annual Simulation Symp, 2007

Topics in Data Mining/Machine Learning

• Use of unlabeled data to estimate class priors.
• Training set data often expensive to collect.
• So, have some data unlabeled we show that even
  this data helps estimate priors.
• Submitted to 2007 International Conf. on Machine
  Learning.
• Novel solutions to the multiple inference problem
  in rule discovery. (In progress.)

47
Biomedical Data Visualization
Contact Kwan-Liu Ma ma_at_cs.ucdavis.edu
http//www.cs.ucdavis.edu/ma
48
Visualization for Homeland Security

• Visual analysis of terrorist network

Contact Kwan-Liu Ma ma_at_cs.ucdavis.edu
http//www.cs.ucdavis.edu/ma
49
Visualization for Cyber Security

• Characterizing network scans uses wavelet
  analysis, visualization, and machine learning.

Contact Kwan-Liu Ma ma_at_cs.ucdavis.edu
http//www.cs.ucdavis.edu/ma
50
Visualization for Cyber Security

• Visual analysis of network traffic

Contact Kwan-Liu Ma ma_at_cs.ucdavis.edu
http//www.cs.ucdavis.edu/ma
51
Visualization for Cyber Security

• Visual analysis of network connectivity

Contact Kwan-Liu Ma ma_at_cs.ucdavis.edu
http//www.cs.ucdavis.edu/ma
52
Visualization for Engineering Design

• Interactive visualization of time-varying 3D flow
  fields
• Collaborative data analysis
• Novel interface design

Contact Kwan-Liu Ma ma_at_cs.ucdavis.edu
http//www.cs.ucdavis.edu/ma
53
Software Project Visualization

• Visualizing the evolution of the software,
  developer email network, file/directory access,
  source code, and their interrelationships.

Contact Kwan-Liu Ma ma_at_cs.ucdavis.edu
http//www.cs.ucdavis.edu/ma
54
Embedded Systems Computer Architecture

• Virtualization Workload Migration for power,
  performance, durability
• Flexible Computation - trading resources and
  quality of results or accuracy of computation
• Synchroscalar Tile based architectures for low
  power and high performance computing
• Scalable and high-performance FPGA architectures
• Please contact Venkatesh Akella
  (akella_at_ucdavis.edu)

55
Professor Richard Spencer

• Research interests Analog and mixed-signal
  circuits for baseband and RF communication.
  Especially low-noise and low-distortion circuits
• Recent and current projects
• Timing recovery for very-low SNR partial-response
  channels and 10 Gb/s backplanes
• Analog Front-end design for Gigabit Ethernet
• Highly efficient, linear, CMOS RF power
  amplifiers
• CMOS implementations of impulse radios
• Linearization of High-Frequency Communication
  Circuits

56
John Owens

• Broad interest Commodity Data-Parallel
  Processors
• How to build, how to program
• Recent research General-Purpose Computation on
  Graphics Processors (GPGPU)
• GPUs have 10x performance vs. CPUs, but are
  difficult to program and have a restrictive
  programming model
• Solution Tools, abstractions, algorithms for
  CPU-GPU application development
• Challenges programming models, partitioning
  between CPU/GPU, data-parallel libraries and
  abstractions
• Partners NVIDIA, ATI/AMD, LANL, Pixar

57
Interconnected SONET/ SDH Rings

• PI Biswanath Mukherjee
• PhD students Smita Rai, Lei Song
• Sponsor Fujitsu Research Labs, Sunnyvale (and UC
  MICRO)
• Dual-node interconnection with drop-and-continue
  de-facto standard for interworking ring
  protection architectures
• Explore provisioning algorithms respecting
  time-slot and alignment constraints
• Minimize capacity fragmentation

58
Metro Ethernet NetworksEnhancing Performance and
Quality of Service
Motivations

• Metro Ethernet Advantages
• Cost effectiveness
• Flexibility
• Rapid provision on demand
• Ease of internetworking

• Ethernets Drawbacks
• Poor Resilience
• No Load Balancing Mechanism
• Simple QoS Scheme
• No Security Mechanism

Concept
Objectives

• Allowing a flow to be on multiple Spanning Trees
  simultaneously
• Prior to a faulty/congested link, COST elevates
  the traffic to the next higher Spanning Tree (ST)
  in the sequence of ST

• Reduce Reconvergence Latency
• Distribute Traffic Loads Evenly Across Network
• Service Differentiation
• Traffic Policing

Contributions

• Keep flows uninterrupted
• Dynamic load balancing
• Increase link utilization
• Protect high priority traffic in the face of
  failure and high congestion period

For more information contact Prasant Mohapatra
(prasant_at_cs.ucdavis.edu)
59
Large-Scale Failures in the Internet

• Approach
• Generation of Internet-like topologies
• Simulation of large scale failures
• Modification of the Border Gateway Protocol to
  accelerate the convergence after a failure

• Objective
• Characterization
• Effect of major factors
• Identification of appropriate metrics
• Designing new routing algorithms to reduce the
  impact of failures

• Inferences
• Degree distribution has big impact on convergence
  delay after failure
• Convergence delays can be improved if updates are
  sent at a quicker rate
• But adaptive schemes are needed to control the
  number of updates
• Updates can be analyzed to quickly identify
  regions of failure and to avoid problematic
  routes

For more information contact Prasant Mohapatra
(prasant_at_cs.ucdavis.edu)
60
Competitive business strategies for information
technologies
Contact Hemant Bhargava, hemantb_at_ucdavis.edu,
Graduate School of Management, UC Davis
61
For further information

• If you need additional information about any of
  the projects, please feel free to contact the
  professors directly. The slides have the contact
  information.
• For any other questions, comments, or if you are
  looking for additional expertise or information
  regarding CFIT, please contact
• Prasant Mohapatra (prasant_at_cs.ucdavis.edu)
• John Vicente (john.vicente_at_intel.com)