wireless by 06nZI45

VIEWS: 0 PAGES: 90

									Topics in Wireless
    Networks
      Robert Turetsky
   rjt72@columbia.edu
   Columbia University
        12/6/2000
        Presentation Goals
• Introduce major wireless technologies
  – Wireless Application Protocol (WAP)
  – Metricom: Ricochet
  – Sun Microsystems: JINI
  – Bluetooth
• Identify issues in wireless network design
• Discuss recent research in wireless
  data/computer networks
            Why Wireless?
•   Freedom from the boundaries of space
•   Instant access to data and applications
•   Spontaneous (ad-hoc) networking
•   Embedded communication in devices
•   M-Commerce
•   Play games and find stock quotes
    during lame status meetings
       So, what’s the holdup?
•   High communications, device cost
•   Long latency period
•   Low/uncertain bandwidth
•   Unreliable channels
•   Roaming/handoff issues
•   Unique devices
•   No accepted standards
•   We want it NOW issues
     Designing for wireless
• Lacks non-commercial testing grounds
• Goal: Minimize download time, manual
  typing
• Content bytes
• Service oriented applications
• myCustomization
• Interaction with ‘wired’ web
• Always consider different screen sizes
    The present state of wireless
•   WAP: Web services on cellular phones
•   iMode: M-commerce overseas
•   Ricochet: 128 kbps wireless modems
•   Connected PDAs / E-mail devices
•   Transmeta Caruso: Low-cost/power P
•   Voice/Data network separate (WAP)
•   Data over voice (cellular modems)
WAP: Wireless Application
       Protocol
        WAP: Introduction
• WAP: Wireless Application Protocol
• Communications Protocol and
  Application Environment for deployment
  of information resources, advanced
  telephony services and Internet access
  from mobile devices
• WAP Forum: Phone.com, Ericsson,
  Nokia and Motorola
• Check out my cool phone!
         WAP: Philosophy
• Simplicity in devices, complexity in
  gateways
• Base standards on wired web
• Designed for any mobile phone
• Designed for any wireless protocol
          WAP: Applications
•   Wireless Web Browsing, E-mail
•   Remote LAN access
•   Mobile Banking
•   M-Commerce and prepayment
•   Remote point of sale
•   Wired workers, ie meter readers
•   In the future: Location-based services
•   In the future: Multimedia
        WAP: Architecture




• WAP Gateway acts as intermediary
  between web server and mobile client
         WAP: Terminology
•   WAP Device
•   WAP Client
•   Microbrowser
•   User Agent
•   WAP Gateway
•   Network Operator
•   Bearer Services
•   Origin Server
   WAP: The Protocol Stack




• Mobile enhancements to OSI Protocol Stack
 WAP WAE: Application Layer
• Provides an environment to execute and
  develop applications
• Building blocks:
  – WML: Wireless Markup Language
  – WLMScript: Lightweight scripting language
  – 2 User Agents:
    • WAE: Microbrowser and text editor
    • WTA: Wireless Telephony Agent (in
      development)
WAP: WML Markup Language
•   Optimized for narrowband pipes
•   Based on early stages of HTML
•   Defined as an XML document type
•   Basic unit: The Deck
    – Made up of one or more Cards
    – One deck downloaded at a time
    – Variable exchange between cards
    – How to display a deck is up to the browser
        WAP: WML Supports:
•   Text editing and text formatting
•   Bitmapped b/w images (WBMP)
•   User input elements
•   Variables
•   Navigation and history stack
•   Unicode
WAP: WML Example
       <?xml version="1.0"?>
       <!DOCTYPE wml PUBLIC "-
         //WAPFORUM//DTD WML
         1.1//EN"
         "http://www.wapforum.or
         g/DTD/wml_1.1.xml">

       <wml>
         <card id="main" title =
         "First Example">
           <p> Hello world! </p>
         </card>
       </wml>
         WAP: WMLScript
• Based on EMCAScript
• Adds intelligence to the client:
  – Math and string functions
  – User input validation
  – Access to device facilities
  – Methods for interacting with user
• WMLScript files cannot be embedded in
  WML pages like JavaScript.
       WAP: Session Layer
• Session services: set up a communication
  channel between client and server
• Shares state between client and server
• Binary form of HTTP/1.1
• Two protocols like TCP and UDP:
  – Connection oriented, runs over WTP
  – Connectionless, runs over WDP
    WAP: Transaction Layer
• WTP: Wireless Transaction Protocol
• Provides services to accomplish
  transactions with varying degree of
  reliability
• Operates over WDP or WTLS
  WAP: Transaction reliability
• Unreliable request   • Reliable request




• Reliable request with result message
       WAP: Security layer
• Based on HTTP security layers
• Meant to encourage development/use
  of sensitive wireless applications
• Ensures:
  – Privacy
  – Server Authentication
  – Client Authentication
  – Data Integrity
          WAP: Conclusions
•   WAP extends the WWW to wireless
•   WAP is still in its infancy
•   WAP detractors abound
•   3G will provide new services and a test
•   Patience is a virtue, but time is of the
    essence
 RICOCHET
Wireless High-Speed WAN
                   The Ricochet Network
                       Infrastructure
                                             NOC         NOC



             NIF                                   NIF                                     NIF


WAP      WAP              WAP          WAP        WAP          WAP         WAP         WAP              WAP


 Microcell         Microcell   Microcell   Microcell   Microcell   Microcell   Microcell         Microcell

 Microcell         Microcell   Microcell   Microcell   Microcell   Microcell   Microcell         Microcell

 Microcell         Microcell   Microcell   Microcell   Microcell   Microcell   Microcell         Microcell
    Ricochet Network Components

• Wireless Radio Modems
• Microcell Radios
• Wired Access Points (WAP)
• Regional Network Interface Facilities
  (NIFs)
• 2 Network Operations Center (NOC)
Ricochet Infrastructure Take 2
Ricochet Technology Overview
•   Licensed 2.3GHz and Unlicensed ISM bands
•   Packet switched to shared bandwidth
•   Frequency hopping for multiplexing
•   Microcells meshed together:
I DREAM OF JINI

JINI: Sun Microsystems
               JINI: Goals
• Enabling users to share resources and
  services across a network
• Providing users easy access to resources
  anywhere on a network while allowing the
  network location of a user to change
• Simplifying the task of building, maintaining
  and altering a network of devices, software
  and users
    Jini System Components
• A set of components that provides an
  infrastructure for federating services in a
  distributed system
• A programming model that encourages the
  production of reliable distributed services
• Services that can be made part of a federated
  Jini system which offer functionality to any
  other member of the federation
• End Result: A network that allows shared
  services to join and detach from a network
         Jini and Wireless
• Jini can be run over any network that:
  – Has inter-service communication channels
  – Supports the JVM
• Resource Discovery allows for
  embedded devices and ad-hoc
  networking
            Java and Jini
• Jini extends the JVM from one machine
  to a distributed network
• Why use Java?
  – Bytecode and Data cross-platform
  – Strong typing
  – Java Security Manager
  – Java RMI as framework for distributed
    object communication
  – Because Sun Micro likes Java
           JINI: Definitions
•   Service
•   Federation
•   Service Protocol
•   Lookup Service
•   Discovery
•   Lease
•   Transaction
•   Event
 JINI Components: Infrastructure

• Discovery/join protocol
• Lookup services
• Distributed security system
  – Built into Java RMI
  – Principal: agent accessing a service
  – Access Control List
JINI Components: Programming
           Model

• Set of interfaces to be implemented
  by services:
  – Leasing Interface
  – Event Notification Interface
  – Transaction Interface
    • 1) Voting Stage
    • 2) Coordinated Commit Stage
  JINI Components: Services
• Can be computation, storage,
  communication link, I/O devices, special
  hardware or another user
• Appear as objects written in Java
  – Others don’t care if service is software,
    hardware, on different machine,
    shared/distributed memory space
  – Public interface for other services to access
  – One service can be built of multiple services
           JINI Protocols
• Discovery: A service looks for a Lookup
  Service to register with
• Join: Connecting a service with a
  Lookup Service
• Lookup: A client or user locates and
  invokes a service
• Discovery/Join takes place on device
  startup
JINI Protocols in action
    KVM: Java Everywhere!
• KVM: Highly optimized VM for resource
  constrained user devices
• Idea: Add ‘profiles’ for optional features
• Implemented as 16,000 Lines of C
• Includes subset of java.lang and
  java.util, but no AWT
• Take it for a test drive!
ROB TO BRIDGE:
BLUETOOTH
  The future is now….
.…whatever that means.
 Bluetooth: What does it do?
• Facilitate fast and secure wireless
  communications
• Replace cables for everything
• Handle voice/data communication for
  spontaneous (ad-hoc) networks
• Provide a medium for embedded
  devices to communicate
Bluetooth: Potential applications
•   Syncing multiple devices
•   Creating Spontaneous LANs
•   Building Internet Bridges
•   Smart appliances (embedded everything)
•   Three-in-one phones
•   Instant peer-to-peer document exchange
•   Making geeks rich again
       Bluetooth: The numbers
•   Transmits up to 30 ft
•   2.4GHz frequency band (ISM)
•   Data transfer rate: near 1Mb/s
•   Duplex communication:
    – Asymmetric: 721kb/s one way, 57.6kb/s
      other way
    – Symmetric: 423.6k/s
 Bluetooth: More Numbers
• Combination Circuit / Packet Switching
  – Circuit switching: 1 packet send, 1 packet receive
  – Packet Switching: 1600 hops/sec, 1, 3 or 5 slot packets
• Power Consumption: 3% of Cellular phones
• Synchronous/Asynchronous communication
  – 1 Asynchronous packet switched Data channel
  – 3 Synchronous circuit switched voice channels
  – 1 shared data/voice channel
• Receiver interconnections: point to point and
  multipoint
      Bluetooth: Data Integrity
• Works under heavy radio noise/interference
  – Fast hops with short packets
  – Switches frequency after each packet
• Error Correction Code: FEC
  – Protects against noise from long distance
  – Protects against interference from microwaves
• Built in encryption/authentication schemes
• 09/2000: Lucent researches cracked
  security codes
       Bluetooth: Piconets
• Piconets: A collection of bluetooth
  devices connected together in an ad-
  hoc fashion.
• Between 2-8 devices per piconet
• One device in the piconet acts as the
  master the rest are slaves.
  – Any device can be the master
• Connect piconets together to form
  scatternets
Bluetooth: Access Procedures




• Use page and inquiry protocols to
  establish and maintain connections
  between devices
    Bluetooth: Link Manager
• Responsible for the following services:
  – Sending and receiving data
  – Name request
  – Link address inquiries
  – Connection setup
  – Authentication
  – Link mode (data/voice) negotiation
• Discovers and communicates with other
  LMs via Link Manager Protocol (LMC)
    Bluetooth: Connection Mode




• In connection mode, packets can be sent from the
  master to slave devices.
• 4 modes of operation: Active, Sniff, Hold and Park
        Bluetooth: Packets




• Access code can be: CAC, DAC, IAC
• Packet header has the following fields:
  AM_ADDR, FLOW, SEQN, TYPE, ARQN, HEC
• Payload can be voice, data or both
    Bluetooth: No Handoffs?
• Raison D'être: Cable, not cellular
  replacement
• To keep costs <$5 means fast, efficient
  and small
• Piconets are dynamic and fast
  – Can connect to a new piconet in 1-5s
• Solutions do exist, but they are
  proprietary (and incompatible)
          Bluetooth vs. Jini
• Different layers of the protocol stack:
  – Bluetooth: Physical and transport layer
  – Jini: Resource discovery protocol
• Different approaches to device discovery
  – Bluetooth: Simple and fast
  – Jini: Robust
• Jini over Bluetooth?
  – Not yet. Jini is TCP/IP based. Bluetooth is
    not. But it is in the works :)
          Bluetooth: SIG
• Promoter companies: 3Com, Ericsson,
  IBM, Intel, Lucent, Microsoft, Motorola,
  Nokia and Toshiba
• SIG members can develop (royalty-free)
  Bluetooth devices and products
• Any incorporated company can join for
  free
• 2000 member companies
      Cool Bluetooth Products
•   National Semiconductor demos
•   Ericsson Bluetooth cellular phone
•   Motorola + others: Bluetooth cards
•   Connected Library
•   Lots more to come….
WIRELESS PAPER
    SURVEY
   1994 to 2000
        Paper Survey: Topics
•   Wireless Andrew
•   MobileIP and IEEE 802.11
•   A Cordless Distributed System
•   Optimizing HTTP for wireless
•   Predictive Mobility Management
•   Routing Protocol Evaluation
  Wireless Andrew (CMU 1997)
• Deals with issues in setting up a campus-
  wide wireless WAN
  – Infrastructure choices
  – Business decisions
  – Which technology to use
  – Dealing with holes in service, interference
  – Manageability, maintainability of the network
  – Comparison to the ‘wired’ Andrew file system
  Mobile IP and IEEE 802.11
• Network layer solution for wireless LANs
• Main Idea: Forward mobile user’s
  location information to home base
• 2 types of network configuration:
  – Ad-hoc
  – Infrastructure
• Good: Works with IPv4
• Bad: Doesn’t work with IPv6
         Mobile IP Networks
• Ad-hoc
  – Usually fully-connected network
  – Ways of discovering peers:
     • SEA: Spokesman Election Algorithm
     • Flood (broadcast) to anyone who can listen
• Infrastructure
  – Fixed access points
     • Can be easily connected to ‘wired’ networks
     • Handoff at overlaps between access points
 IEEE 802.11 Physical Layer
• 2.4 GHz frequency band (ISM)
  – Direct sequence spread spectrum
  – frequency hopping spread spectrum
• 300-428,000 GHz (Infrared)
  – Bad: Line of sight only
  – Good: More secure
    802.11 MAC Layer Spex
• MAC Layer: Medium Access Control
• Responsible for collision avoidance
  – CSMA/CA: Carrier Sense Multiple Access
    with collision avoidance
    • Check to make sure a node is free
    • If free, transmit. Else random backoff
  – Ethernet’s collision detection scheme can
    not be used.
802.11 Hidden Node Problem
802.11 Handshaking Protocol
• Sender sends RTS (ready to send)
  packet with CONTENT_LENGTH
• Receiver replies with CTS (clear to
  send) packet
• Sender sends packet + CRC
• Receiver checks CRC. If OK, confirm
  with an ACK
BlueSky: Cordless Net for Palmtops
• Protoype Personal Area cordless network
  – Built from cordless phone radio modules
• Mobile users plug adapter into serial ports
• AccessPoint: Bridge to wired nets
• Uses PPP but keeps connections open
  during handoffs
• Can have 10-15 clients sharing 150kbps
• IBM, Bourns College of Engineering: 1999
  BlueSky: Properties of a Good
       Cordless Network
• Low Power consumption
  – Large number of short-range Access Points
  – Forget packet-based wireless LANs (ie 802.11)
• Low Control traffic
• Security: Authentication at each handoff
• Low Connection-related overhead (ie not
  802.11)
     BlueSky: Access Points
                    • Responsible for:
                      – Sharing bandwidth
                      – Registering new
                        devices
                      – Performing handoffs

• One communication channel per AP
• Neighboring APs use different channels
  to avoid interference
BlueSky: Cordless Modules
           • Why Cordless?
             – Cheap and mass-
               produced
             – Circuit-based
             – Low Power
               Consumption
        BlueSky Layers: PPP
• BlueSky uses PPP as its communications
  protocol
  – Everyone already has it on their
    laptops/palmtops
  – Can leverage existing dialup programs (ie
    Dialup Networking for Win98/2000)
  – Both Serial line and cordless link are circuit
    oriented
  – Can easily tunnel packets to wired networks
     BlueSky Layers: MAC
• Packet Based for multiplexing
• One device can communicate at a time
• All other devices must turn of
  transmitters
• Polling-Based (moves complexity from
  device to Access Point)
 WebExpress: Wireless Web
   Browsing (IBM 1996)
• Idea: Improve HTTP performance over
  wireless networks
• Inhibitors to wireless HTTP:
  – High Latency (Oracle: 15s for TCP/IP
    request)
  – High Cost (1996: $1 per 10k)
  – Low bandwidth (4800bps-19.2kbps)
  – High unreliability
WebExpress: More HTTP Woes
• Connection overhead: must open new
  connection for every HTTP object
• Basically Stateless: Browsers must re-
  send their capabilities
• No support for caching dynamic
  documents
• HTTP is verbose
   WebExpress: The main idea
• WebExpress Intercept Layer
  – Runs between wireless link and HTTP
  – Client Side (CSI) and Server Side (SSI)
    Intercepts
  – No need to reconfigure browsers or web server
 WebExpress: Optimizations
• Caching: on server and client side
• Differencing: cache dynamic documents
  and only transmit differences
• Protocol Reduction: a CSI maintains its
  TCP/IP connection to the SSI for all
  HTTP requests.
• Header Reduction: transmit browser
  capabilities only once
      WebExpress: Caching
• Differences between wired and wireless:
  – Cost of communication
  – Persistence
  – Applicability
  – Scope
         WebExpress: Cache
           Improvelments

•   Cross-session cache persistance
•   Age-based coherency algorithms
•   User-configurable options
•   Digital-signature based modification
    verification
WebExpress: Differencing
   Mobile Distributed System
          Architecure
• Ericsson and Swedish Inst. Of
  Technology, 1995
• Main Idea: To improve performance and
  mobility of wireless distributed systems
  – Location sensitive information management
  – Use agents to track and predict user mobility
  – Forward user state to other agents to
    minimize handoff time delays, traffic
TLAs: Three Letter Acronyms
• MDSP: Mobile Distributed System
  Protocol
• LSIM: Location Sensitive Information
  Management
• PMM: Predictive Mobility Management
  (also called MMP)
• Agent: a space on Access Point that is
  synchronized with mobile user’s file
  system, memory space and cache
    Agents: The Main Idea




• Use mobility prediction to have data
  ready for a user once he arrives in a cell
 Three Not-so-secret Agents
• M-Agent: representative of the mobile
  user to the network
  – Creates and deletes MF-Agents
  – M-Agent the same through entire session
• MF-Agent: Mobile floating
  – Temporary agent set up before user’s
    arrival into a cell.
  – Becomes AM-Agent when user arrives
• AM-Agent: Active agent
  – delivers resources to the user
  Pre-assignment procedure
• Mobile terminal sends assignment
  request to M-Agent or AM-agent
• M-Agent registers the request and
  forwards it to the MSDP base
• MSDP tries to create new MF-Agent
• New MF-Agent asks foreign agent if it is
  OK to reside there
     Pre-assignment Cont’d
• If OK, new MF-Agent sends mobile
  terminal a success message
• The new MF-Agent will maintain
  variable data consistency with the AM-
  Agent
• Now, MF-Agent will be ready for the
  mobile agent’s arrival
• If MF-Agent creation fails, an error
  message is sent.
     MMP: Mobility Prediction
• Goal: To have an MF-Agent waiting for
  the mobile user without wasting resources
• The proposal: Predict where the user is
  likely to go, and set up MF-Agents there.
• 2 types of movement:
  – Regular: Can be predicted
  – Random: Cannot be predicted
MC/MT/d assignment protocol
• Tries to predict two types of movement:
  – MC: Movement Cycle (cyclical)
  – MT: Movement Track (unidirectional)
• Benefits of MC/MT/d
  – Less overhead than assigning MF-Agents
    to all neighbors (PTP protocol)
  – Avoids computation-intensive stochastic
    models
  – Prediction accuracy of 95%
         MF/MC/d method
• Calculate mobility during m
• Define circle of radius d = floor(hmm)
  – h = hierarchic factor
  – m = mobility density during time m
• Assign MF-Agents to MC/MT predicted
  cells within radius d
• Repeat when user reaches the
  boundary of the circle
   MC/MF/d: The Bottom Line
• Compare MC/MF/d to:
  – Standard: Without MF-Agents
  – PTP: Assign MF-Agents to all cells within d
• Performance = reduced latency -
  increased overhead
• PTP has performance gain of 25% over
  Standard
• MC/MF/d performance gains:
  – Heavy load, good prediction: 45%+
  – Low load, bad prediction: 65%
Routing Algorithms Evaluated
• Uses simulation to assess performance of
  various routing algorithms
• Link-state protocol
• Distance Vector protocols
  – Extended Bellman-Ford
  – Destination Sequenced Distance Vector (DSDV)
• TORA: Multipath protocol
• On-demand protocols
  – DSR
  – AODV
  Routing protocol evaluation
• Proactive approaches led to low delays
  and high delivery rates, but increased
  traffic
• On-demand protocol led to worse
  delays and delivery rates, but reduced
  load
• TORA did not perform well
• As user count increased, routing traffic
  varied less and less among protocols
        Wireless Networks:
            Conclusion
• Bill Joy’s six webs: Wireless is one key
  to the connected future
• This is only the beginning…..

• You can reach me at
  rjt72@columbia.edu
• A copy of this presentation is at
  http://www.caip.rutgers.edu/~robjt/wireless.ppt

								
To top