Towards Wireless Overlay Network Architectures

advertisement
The Bay Area Research Wireless
Access Network (BARWAN)
Low-tier
New Ideas
• Wireless Overlay Internetworking
–
–
–
–
–
Satellite
Regional Area
High-tier
“Overlay” IP extensions to Mobile IP
Low Latency Inter-subnet Handoffs
High Thruput Reliable Transport
Class-Based Queuing Link Management
Subnet Bandwidth Load Balancing
• Client-Proxy-Server Architecture
–
–
–
–
–
Local Area
Wide Area
High Mobility
Low Mobility
R. H. Katz & E. A. Brewer, UC Berkeley
Subcontractor: Hughes Research Labs
Impact
• Fundamental technology for 21st century
battlefield communications: support for wide
diversity of hybrid & asymmetric link
technologies, and end device display &
computation capabilities
• Seamless roaming & application adaptation
across 3–4 orders of magnitude of wireless b/w
and latency (10 kbps to 10 mbps, 1 ms to 1 s)
• Demonstrate network & application techniques
able to scale to support 10s of data users/room,
100s/building, 1000s/facility, 10000s/metro, and
100000s/region
Bandwidth-aware Data Type Adaptation
Web & A/V Data Types over Wireless
Proxy-Aware API, Kerberos Integration
Delivery Class Abstraction
Scalable Wireline Processing for Mobiles
Schedule
Measure &
Eval Wireless
Ovrly Technologies
Aug 95
Start
Demo in-building
ovrlys & h/os with
dynamic b/w alloc
Aug 96
Demo wide-area ovrlys
& low latency h/os w/
subnet load balancing
Aug 97
Early Prototype
Proxies (1-10 users)
Proto Scaled Proxies
(100-1000 Users)
Initial Architectural
Specification
Scaled Architectural
Specification
Aug 98
End
Scaled Proxies
(1000-10000 Users)
Large Scale
Scalability
Demonstrations
1
The Bay Area Research Wireless Access Network:
Towards a Wireless Overlay Internetworking Architecture
Satellite
Regional Area
Low-tier
High-tier
Local Area
Wide Area
High Mobility
Low Mobility
Randy H. Katz and Eric A. Brewer
Computer Science Division, EECS Department
University of California, Berkeley, CA 94720-1776
Subcontractor: Hughes Malibu Research Laboratories
2
Presentation Outline
•
•
•
•
•
•
•
Retreat Purpose and Agenda
Project Objectives, Motivation, and Approach
Project and Testbed Status
Technology Developments
Review Project Plan
New Directions
Summary and Conclusions
3
Presentation Outline
•
•
•
•
•
•
Retreat Purpose and Agenda
Project Objectives, Motivation, and Approach
Project and Testbed Status
Technology Developments
Review Project Plan
New Directions
4
Retreat Goals &
Technology Transfer
People
Project Status
Work in Progress
Prototype Technology
Early Access to Technology
Promising Directions
UC Berkeley Project Team
Industrial Collaborators
Reality Check
Hughes Researchers
Government Sponsors
Feedback
Friends
5
BARWAN/Hughes Team
• Networking
• Applications
– Hari Balakrishnan (Reliable Transport)
– Tom Henderson (Satellite-based
Transport & Routing Protocols)
– Todd Hodes (Mobile Location Services)
– Giao Nguyen (Channel Scaling)
– Venkat Padmanabhan (Sessions•
Transport Interface)
– Mark Stemm (Performance Discovery)
– Helen Wang (Vertical Handoff)
• Hughes Malibu Research Laboratory
– Son Dao
– Yongguang Zhang
– Dante De Lucia
– Elan Amir (Media Gateway/MASH)
– Yatin Chawathe (Proxy
Architecture/MASH)
– Armando Fox (Proxy Architecture)
– Steve Gribble (Scalable Servers)
Technical Support
– Brian Shiratsuki (System Admin)
– Keith Sklower (Sys Programming)
• Admin Support
– Terry Lessard Smith
– Bob Miller
Tao Ye finished her MS and went to JavaSoft
Daniel Jiang went to Daimler Benz Research Laboratory
6
BARWAN Sponsors and Participants
• DARPA GloMo Program
–
–
–
–
Rob Ruth, DARPA PM
US Army CECOM
NIST
SRI
• Industrial Supporters
–
–
–
–
–
–
–
Daimler Benz
Ericsson (Reiner Ludwig, Visiting Industrial Fellow)
Fuji Xerox Palo Alto Labs
Hughes Research
IBM
Metricom
Toshiba (Masahiro Takagi, Visiting Industrial Fellow)
• Friends
– Hybrid Networks, Packeteer, Sandia, Stanford
7
Retreat Schedule
• Wednesday, January 14:
1200 - 1300
1300 - 1400
1400 - 1500
Fox
1500 - 1530
1530 - 1800
»
»
»
»
»
Lunch
Project Overview and Status, Randy Katz
What We Learned About Extensible Proxies, Armando
Break
Research Highlights
Passive Network Performance Discovery - Mark Stemm
Transport over Satellites - Tom Henderson
Multicast over DBS - Yongguang Zhang
OS Support for Scalable Network Services - Steve Gribble
Mobility Enhanced Network Services - Todd Hodes
1800 - 1930 Dinner
1930 - 2100 Poster Session and Demos
2100 Distributed Interactive Collaboration (aka Riven)
8
Retreat Schedule
• Thursday, January 15:
0730 - 0830 Breakfast
0830 - 1000 New Directions, Anthony Joseph
» ProActive Infrastructure, Eric Brewer
» Beyond Third Generation, Randy Katz
1000 - 1030 Break
1030 - 1200 New Directions, Continued
» Brainstorming with sponsors on new directions
1200 - 1300
1300 - 1630
1800 - 1930
1930 - 2100
Lunch
Ski Break
Dinner
Invited Talks
» Transport over Cellular, Reiner Ludwig, Ericsson
» Rover Toolkit, Anthony Joseph, UC Berkeley
» Hybrid Network’s Technology, Subir Varma
2100 -
Distributed Problem Solving (aka Riven)
9
Retreat Schedule
• Friday, January 16:
0730 - 0830
0830 - 1000
1000 - 1030
1030 - 1200
1200 - 1300
1300 -
Breakfast
Six Month Planning Session, Eric Brewer
Break & Check-out
Sponsor Feedback Session, Randy Katz
Lunch
Depart Granlibakkan
10
Presentation Outline
•
•
•
•
•
•
•
Retreat Purpose and Agenda
Project Objectives, Motivation, and Approach
Project and Testbed Status
Technology Developments
Review Project Plan
New Directions
Summary and Conclusions
11
Heterogeneous Mobile Computing
“People and their machines should be able to access information and communicate
with each other easily and securely, in any medium or combination of media -voice, data, image, video, or multimedia -- any time, anywhere, in a timely,
cost-effective way.” G. Heilmeier, 1992
•
•
Access
Anytime, Anywhere
– Wide-Area Coverage
– Scalable Processing
– Highly Available Operation
•
• Three Overarching Strategies:
– Heterogeneous Wireless Networks
– Network Optimization
– Dynamic Adaptation
Easily
– Transparent Access
– Localized Service
•
Securely
– Global Authentication
•
Any Medium
– Multimedia: Audio/Video/Graphics
•
Timely
– Performance
•
Cost Effective
– Heterogeneous Support via Proxies
12
Wireless Overlay Networks
Regional-Area
Theatre of Operations
Metropolitan-Area
Rear Echelons
Campus-Area Packet Relay
Bases, Depots, Ranges
In-Building
Command Centers
Training Centers
13
Asymmetric & Heterogeneous Access
High Bandwidth
– Command Post
– Disaster Relief
– Remote Clinic
– Organization w/poor
Internet connectivity
Low Bandwidth
Local
Subnet
14
XXI Century Battlefield
Architecture
Rear Area
Battlefield
ATM Backbone
Packet Radio Network
Radio Access
Point
15
BARWAN Testbed
DirecPC DBS
(1 mbps)
Vertical
Hand-Off
LOS Wireless Cable
(6 mbps)
DARTNet II
CAIRN
10-30 kbps
Metricom
Cellular Modem
Cellular Packet Data
Packet Radio
16
Client-PROXY-Server
Architecture
• Proxy
–
–
–
–
Mediates between wireless and wireline environment
Ideally executes at “well-connected” boundary of internetwork
Manages caches and chooses transport data representations on-the-fly
Trade transcoding time against communications time
Well Connected
Proxy
Poorly
Connected
17
Overlay Network Challenge
Type of
Network
In-Building
Campus-Area
Packet Relay
Network
Metro-Area
(Wireless
Cable)
Wide-Area
Bandwidth
>> 1 Mbps
Comm’l RF: 2 Mbps
Research IR: 50 Mbps
- 64 Kbps
Latency Mobility
< 3 ms
Pedestrian 2-Way ’ractive
Full Frame Rate
(Comp)
- 100 ms Pedestrian
< 10 ms
19.2 Kbps
High Quality
16-bit Samples
22 KHz Rate
Med. Quality
Slow Scan
Med. Quality
Reduced Rate
> 100 ms
Stationary 2-Way ' ractive
Full Frame Rate
(Compressed)
Vehicular Freeze Frame
High Quality
16-bit Samples
22 KHz Rate
Asynchronous
“Voice Mail”
4.8 kbps–10+ Mbps > 100 ms
(LEO/DBS/VSAT)
(asymmetric)
Vehicular Seconds/Frame
Stationary Freeze Frame
Asynchronous
“Voice Mail”
Regional-Area
10-30 mbps
(one way, LOS)
Typ Video
Typ Audio
Performance Performance
Latency as critical as bandwidth in wireless networks
Wide diversity of network performance parameters
Competing infrastructure providers
Pedestrian vs. vehicular mobility
18
Application Support Challenge
Device
High-end
PC
Low-end
PC
High-end
notebook
Low-end
notebook
PDA
Bandwidth,
bits/sec
Ethernet
(10Mbits),
ISDN (128K)
CPU
266 Mhz
Pentium Pro
150 Mhz
Pentium
Mem/
Disk
64/4G
16/1G
Cellular (9600)
or wireline
(28.8K) modem 100 Mhz 486
2400-14.4K
modem
20+ Mhz
RISC or x86
2/0
Screen
Bits/
size
pixel
1280x1024 16-24,
color
1024x768 8-16,
color
800x600
8, color
640x480
4, gray
320x200
1–2, gray
Client variation spans an order of magnitude
19
Cross-Cutting Architectural Issues
• Dynamic resource allocation and adaptation
– Proxies: adapt representations for available bandwidth and latency
» Vertical handoff-based event notification
» Proxy transcoder load balancing & refinement caching
» Performance discovery
– Network: meeting performance “promises” for classes of users and
types of data
» Choosing “best” available overlay for connectivity
» Scheduling bandwidth-constrained (wireless) links
» Exploit channel state for higher channel utilization
• Leverage existing Internet standards
– Mobile IP, TCP, Service location protocols, HTTP, POP, IMAP, etc.
– But allow architecture-aware applications to obtain enhanced
functionality
20
Daedalus/GloMop Architecture
Basestations
Horizontal
Handoff
Foreign
Agent
Server
(Correspondent Host)
Local
Proxy Host
G
W
GW
G
G W
G W
W
Local
Services
Vertical
Handoff
Proxy Host
IP
Internet
GW
Overlay IP
GW
Wireless Subnets
Home Agent
21
Mobile Applications
“Sessions”
Proxy Agents
Transport
(Asymmetric, Heterogeneous, Lossy Links)
Overlay IP (Mobile IP + Overlays)
Snoop Agent
Link Scheduling
Net Connection
Monitoring;
Net-Appl
Interface;
ELN;
Wireless LAN
Location-Dependent
Services
Daedalus/GloMop Architectural
Components
IP
Wire- DBS
Packet
Less SubRadio
Cable Net
Cellular
Data
ATM
SubNet
Mobile IP
Beacons
22
Presentation Outline
•
•
•
•
•
•
Retreat Purpose and Agenda
Project Objectives, Motivation, and Approach
Project and Testbed Status
Technology Developments
Review Project Plan
New Directions
23
Project Strategy
Architectural Design
Scaled Implementations
Early Prototypes
Proof of Concepts
Measurements & Evaluation
24
Project Plan and Status
67%
Project Start: 15 Aug 95
(Start + 29 Months)
Scalabil
Scalability
95%
Functionality
Demonstrate Ability to
Scale to Large Communities
of Mobile Users
100%
Early
Early Proof
Proof of
of
Concept
Concept
Implementations
Implementations
Establishment of
of
Establishment
BARWAN Testbed
Testbed
BARWAN
100%
Demonstrate Seamless
Roaming over
Local and Wide Area
Measure Alternative
Overlay Network
Performance
NOWs
Local & Wide-Area
Wireless Overlays
Enhanced E2E Performance
Extend with
Emerging
Technologies
25
Achievements—January 1998
• Proxy Development
– Wingman Pilot Graphical Web Browser
– Demonstration of extensible proxies
– MASH Toolkit-TranSend Proxy Server integration
• Overlay Network Development
– Developed new transport protocol tuned for high latency, highly
asymmetric satellite environment (STP)
– Snoop V2 Beta with Explicit Loss Notification
– TCP sessions with integrated congestion control and recovery
– Enhanced “right edge” loss recovery scheme suitable for slow speed
links and short transfer size Web traffic
– Other TCP enhancements for Web traffic characteristics (“fast start”)
• Research Infrastructure
– VINT enhancements for wireless simulation: cable modem network,
LANs, wireless links, asymmetric networks, TCP variations, routers,
RTP
30
Local Area Wireless Testbed
Soda Hall, UC Berkeley
Scalable, High Bandw idth
Low Latency Sw itches
Organized
Fiber Physical
Links
Wireless
Transceivers
Computing / Interactive
“Light” w orkstations
Interactive
Media Access
workstations
Dedicated Computing Resources
~100 “dark” workstations
+ mass store, special servers
• 3rd, 4th, 5th, 6th, 7th floors covered by WaveLAN BSs
• Low latency handoff co-resident with DHCP
• Private Metricom network (1 WAP, 6 Infra Radios)
31
Wide-Area Wireless Testbed
Non-Cooperating Wide-Area
Wireless Networks: no control
over basestations
 Metricom PR Network
 DirecPC DBS Service
 Wireless Cable Network
 Cellular Modems
• Cellular Digital Packet Data
• GSM General Packet Radio
Service
(GSM technology finally
deployed in SF Bay Area
as PCS 1900 by PacBell
Mobile Services in Spring 1997)
32
BARWAN Testbed
GSM Circuit Switched Cellular
Metricom
Wide Area
WLAN RF
Base
IBM IR
Station
DirecPC
Basestation
Hughes DBS
Basestation
Soda Hall
+ Ethernet
Gateway
Metricom "WAP"
Internet
Gateway
33
Presentation Outline
•
•
•
•
•
•
•
Retreat Purpose and Agenda
Project Objectives, Motivation, and Approach
Project and Testbed Status
Technology Developments
Review Project Plan
New Directions
Summary and Conclusions
34
Recent Publications
(since July)
•
•
Seshan, Stemm, Katz, “SPAND: Shared Passive Network Performance Discovery,” USITS '97, Monterey, CA,
December 1997.
Gribble, Brewer, “System Design Issues for Internet Middleware Services: Deductions from a Large
ClientTraces,” USITS’97, Monterey, California, December 1997.
•
Fox, Gribble, Chawathe, Polito, Huang, Ling, Brewer, “Orthogonal Extensions to the WWW User Interface
Using Client-Side Technologies,” Proc. 10th Annual Symposium on User Interface Software and Technology
(UIST 97), October 1997, Banff, Canada.
•
Fox, Gribble, Chawathe, Brewer, “Cluster-Based Scalable Network Services,” Proc. 1997 Symposium on
Operating Systems Principles (SOSP-16), St-Malo, France, (Oct. 1997).
•
•
•
•
•
•
Balakrishnan, Padmanabhan, Seshan, Stemm, Amir, Katz, “TCP Improvements for Heterogeneous
Networks: The Daedalus Approach,” Proc. 35th Annual Allerton Conference on Communication, Control, and
Computing, Urbana, Illinois, October 1997.
Henderson , Katz. “Satellite Transport Protocol: An SSCOP-based Transport Protocol for Datagram Satellite
Networks,” Proceedings of 2nd International Workshop on Satellite-based Information Systems, (WOSBIS
`97), Budapest, Hungary, October 1997.
Hodes, Katz, Servan-Schreiber, Rowe, “Composable Ad-hoc Mobile Services for Universal Interaction,”
Proceedings of 3rd ACM/IEEE MobiCom, Budapest, Hungary, September 1997. Best Student Paper Award.
Balakrishnan, Padmanabhan, Katz, “The Effects of Asymmetry on TCP Performance,” Proceedings of 3rd
ACM/IEEE MobiCom, Budapest, Hungary, September 1997.
Noble, Satyanarayanan, Nguyen, Katz, “Trace-based Mobile Network Emulation,” Proceedings of ACM
SIGCOMM'97, Cannes, France, September 1997.
Stemm, Katz, “Measuring and Reducing Energy Consumption of Network Interfaces in Hand-Held Devices,”
IEICE (Institute of Electronics, Information and Communication Engineers) Transactions on
Communications, Special Issue on Mobile Computing, V. E 80-B, No. 8, (August 1997), pp. 1125-1131.
45
Major Recent Research Activities
• Application Support
– Extensibility of the Proxy Service model
» WingMan Browser, MASH MediaBoard/Proxy integration, many
new aggregation and annotation services
• Overlay Networks
– Satellite Transport Protocol
– Multicast over Satellite (Hughes)
– TCP enhancements: support for asymmetric connections, integrated
congestion control and loss recovery, short transaction transport
connections, session-orientation
• Application-Network Interaction
– Shared Passive Network Performance Discovery
– Mobility Service Architecture and Service Discovery in MASH CoLab
46
Wingman Browser
• Proxy-enabled web browser for
thin clients
– Page layout determined on proxy side
– Full distillation, refinement support
– Thin drawing layer on PDA side
Unwired
Planet
(HDML)
47
Satellite Transport
Protocol
(collaboration with Hughes)
• Most of our work done in the concept of TCP, but
…
• Satellite links are not like WLANs
– Very high latency, well protected link, high bandwidth
• Strategies borrowed from ATM SSCOP
– Poll/Stat handshake
» Receiver explicitly requests retransmissions
» Sender polls receiver for successfully received packets
• Performance
– Throughput about the same as TCP
– Dramatically lower return link b/w demands; insensitive to
variations in round trip delay
– Excellent fit with DBS/asymmetric bandwidth
48
Shared Passive
Performance Discovery
(Collaboration with IBM)
• Passive: piggy back performance
data collection on on-going
accesses, measure bandwidth,
latency, packet loss, etc.
• Shared: share collected information
with other nearby users
• Enables applications to adapt to
network performance in advance
• Running at IBM Watson and
Berkeley
49
Service Discovery
• Adapt device functionality to
services in new environment
– Beacon augmentation
– Adaptive user interfaces
– Composed behaviors
• Deployment within Soda
classrooms and MASH CoLab
Universal Interaction?
– Light, video, slide projector, VCR,
audio receiver, camera, monitor, A/V
switcher control
– Local DNS/NTP/SMTP servers, HTTP
proxies, RTP/multicast gateways
– Audited printer access
– Interactive floor maps, protocols for
advertising object locations
– Coarse-grained user tracking
50
Presentation Outline
•
•
•
•
•
•
•
Retreat Purpose and Agenda
Project Objectives, Motivation, and Approach
Project and Testbed Status
Technology Developments
Review Project Plan
New Directions
Summary and Conclusions
51
Six Month Plan (to Dec 97)
• Documentation
– Finish Scaled Architecture Document (DRAFT available)
– Draft comprehensive system architecture and evaluation paper
• Implementation for Scalability/Extensibility
 Scalable Proxy Architecture
 Demonstrate ease of adding new proxy services
 Extend proxy support for Pilot PDAs
– Implement, Test, Evaluate Strategies for Network Scaling
» Link Scheduling
 New Transport Protocols implemented in PR/Wireless Cable
 Network Connection Monitoring & Delivery Classes
• Extend Testbed
 With Ericsson’s assistance, integrate GSM wide-area connectivity
• GloMo-Wide “Eye Watering Demo”
 Extensive use of proxies, distribution of VHO code
52
Original Research Plan
Year 1
Year 2
Year 3
T1: Overlay Inter network
Management Ser vices
M easure & Eval Ovly Nets Design f or Wide-Area Ovly Demo Wide-Area Ovly
Design f or In-Building Ovly Demo In-Building Ovl y Net Demo Scaled Net Perfor
T2: Mobile Application
Support Services
Design API & App lToolkit Demo API for In-Building
Integrate with NOW Servers
Develop Simple Collab App s Demo Scaled App s Perf
O1: Wide-Area Deployment
and Demonstr ation
Acceler ate Design for Wide- Demo Wide-Area Overla y
Area Ovly Integration
Design f or M ultiple Ovly s
O2: Pilot Application
Demonstr ations
Large Scale Demo of Nav
Deploy In-Build Net@UCSF Design Libr Nav & M ed
Eval M ed Image App s Reqs Image Distr App s using API & M ed Image Distr App s
9/95 - 8/96
9/96 - 8/97
Demo Extension to M ultip le
Wide-Area Overla ys
9/97 - 8/98
53
DARPA GloMo Program Goals
DARPA GloMo
FY 96
Daedalus/BARWAN Program
Adaptive Mobile Internet Services
Measure/eval overlay networking tech
Design overlay network architecture
Design proxy architecture, API, toolkit
Prototype proxies for image, video, maps
Location Transparent Computing
FY 97
Demo B/W Adaptive MM Node
Design Scalable Proxies/Proxy Trans Mgr
Arch for “Remote Collaboration by Proxy”
Demo Advanced Mobile Networking Overlay IP and Vertical Handoff
Reliable transport for hetero/asym nets
FY 98
Demo Multimedia Conferencing
Demo Continuous Mobility
Demo scalable processing for proxies
Demo seamless roaming over in-building,
wide-area wireless overlays
54
Revised Project Schedule
Measure &
Eval Wireless
Ovrly Technologies
Aug 95
Start
Demo in-building
ovrlys & h/os with
dynamic b/w alloc
Aug 96
Demo wide-area ovrlys
& low latency h/os w/
scalable performance
Aug 97
Early Prototype
Proxies (1-10 users)
Proto Scaled Proxies
(100-1000 Users)
Initial Architectural
Specification
Scaled Architectural
Specification
Aug 98
End
Scaled Proxies
(1000-10000* Users)
Large Scale
Scalability
Demonstrations
* on 10 UltraSparc cluster
55
Milestones to End of Contract
• 3Q97
– Final Architectural Specification (slipped to 1Q98)
 Integrated wireless simulation environment based on ns
• 4Q97
– Demonstration of network scaling in WLAN environment
– Demonstration of network scaling in PR environment
• 1Q98
 Demonstration of scalable, extensible TACC (Transformation,
Aggregation, Customization, Caching) Servers
• 2Q98
– Integrated demonstration of wireless networking, real-time
conferencing, proxy adaptation, integrated into GloMo “Eye
Watering Demo”
56
What We Proposed to Do
57
What We Proposed to Do
• Seamless integration of overlay networks
– Handoff
» Low latency handoff via user tracking
» Vertical handoff (power vs. latency)
» Policy-based VHO
– Transport
» TCP over high error rate links (snoop)
• Support services for mobile applications
– Data type specific compression (proxy distillation) and
progressive transmission
– Dynamic applications partitioning across wireless links
(service discovery and adaptation; wingman browser)
– Integration of multimedia and web applications with
wireless environment (vic/vat, browser)
58
What We Proposed to Do
• Managing mobile connections to support
latency-sensitive applications
– Link scheduling, class-based queuing
– Real-time stream adaptation (RTP gateway)
• Load balancing for scalable mobile processing
– Network load balancing across overlay networks
– Network servers to support processing and storageintensive applications (NOW integration/TranSend)
– Uniform architecture for applications support (TACC
programming model )
59
Bonuses Beyond the Proposal
• Research Infrastructure
– VINT-based simulation environment
– Mobile and home IP trace collection
• Asymmetric Transport
– TCP enhancements for bandwidth, latency, error
asymmetries in PR, wireless cable, DBS subnetworks
– Satellite Transport Protocol
• Active Services Architecture
– RTPGateway’s evolution into MediaGateway
– Service discovery and adaptation
• Scalable, Composible Service Architecture
– TACC model
– PalmPilot WingMan Browser
60
Technology Transfer Activities
• Proxy Software
– Beta TranSend binary distribution available
» Distributed to UC Davis, SRI
– Wingman Pilot Browser: 8000 downloads
Postman Pilot E-mail: 6000 downloads
• RTP Gateway
– 171 downloads in calendar 1997
• Transport Layer Software
– 100 snoop V1 downloads in calendar 1997
» Ports to Linux, FreeBSD, NetBSD
– 222 SACK downloads
» Ports to NetBSD
– snoop in daily production use in Reinas wireless network@UCSC
• Wireless, LAN MAC, and Transport simulation modules
distributed to VINT community
61
Presentation Outline
•
•
•
•
•
•
•
Retreat Purpose and Agenda
Project Objectives, Motivation, and Approach
Project and Testbed Status
Technology Developments
Review Project Plan
New Directions
Summary and Conclusions
62
Project Synergies
BARWAN
Wireless Overlay Networks
Scalable Proxies
vic, vat,
wb
RTPGateway
Service Discovery
MASH
Collaboration Applications
Active Services
TranSend/TACC
ProActive Infrastructure
Scalable, Secure Services
Computation in the Network
“Smart Spaces” as an app
Event-Response
Programmable Access
MASH Toolkit
63
ProActive Infrastructure
“Smart Spaces”
Active Routers:
Active network routers
Soft state
Interchangeable
Bases:
Scalable, available servers
Persistent state
Service discovery
Public-key infrastructure
Databases
Home Base
Units:
Client Devices
Sensors & Actuators
User state
E-mail
User tracking
64
ProActive Infrastructure
(Brewer, PI; Culler, Joseph, Katz, Co-PIs)
• The Challenge
– Network-based applications becoming increasingly service intensive
– Computational resources embedded in the switching fabric
– Dealing with heterogeneity, true utility functionality, security, service
discovery
• The Approach
– Build on BARWAN Proxies and MASH Active Services
– Computational Elements
» Units: diverse end devices, from laptops to sensors & actuators
» Active Routers: points of attachment, support for limited
capability units, think “home base station”
» Bases: large centers of storage & computation shared among
many users
– Computational Model
» Connectors, Objects, Functions, Paths
65
Beyond Third Generation
MSC
BSC
VLR
HLR
AUC
EIR
GSM Core
(IP-Based)
GW
Proxy
Server
Next
Generation
Internet
HA
FA
BS
GW
Corporate
Intranetwork
• Convergence of cellular and Internet technology
• Focus on services in telephony context and beyond:
– Mobility, Link Performance Optimization, New Apps and Services
66
Presentation Outline
•
•
•
•
•
•
•
Retreat Purpose and Agenda
Project Objectives, Motivation, and Approach
Project and Testbed Status
Technology Developments
Review Project Plan
New Directions
Summary and Conclusions
67
Summary and Conclusions
• Objective: a complete network and application
support architecture for access across lossy links
from a wide variety of end devices
• Access is the killer app
– Seamless connectivity through wireless overlays
– Adaptivity through proxy services
• Dealing with heterogeneity, asymmetry, adaptation
– Asymmetric bandwidth in satellites, cable modems, cellular
systems: new transport protocol techniques
– High loss links: achieving high bandwidth utilization through local
intelligent retransmission
– Adapt representations to the quality of the end device and its
network connectivity: proxies for audio/video streams and
imageful web documents
• New direction: scalable, available, programmable,
secure services for heterogeneous devices
68
Download