相關資料

國立中山大學視覺與智慧系統實驗室 李宗南教授

 Video Database Systems

沈錳坤、李素瑛﹐視覺資訊系統中內容擷取技術之研究與發展

 Content-Based Image Retrieval (PowerPoint 97)

 Content-Based Video Retrieval (PowerPoint 97)

Data Mining

 Allen (Wei-lun CHANG) Data Mining Research Group

 Mining Market Basket Data Using Share Measures and Characterized Itemsets

 The Fourth International Conference and Exhibition on The Practical Application of Knowledge Discovery and Data Mining      

Data Management in a Wireless Environment

  1. Introduction

    Why do we care about wireless data management?

    • Information is the core of business success.
    • The internet has changed the way information is accessed.
    • Wireless network are in place.
    • Mobile devices are available in the market.
    • Users like to go mobile and get ubiquitous access of information.

    1. Wireless Networks
      • Cellular Systems
        • Analog systems.
        • CDMA digital systems.
          • Data Services: Dial-up, Circuit Switched
          • Coverage areas: 60% of US
          • Roaming: Depend upon carrier
          • Internet access: Yes
          • Type of data support: Dial-up at 14.4 Kbps
        • TDMA digital systems.
          • Data Services: Nothing available today
          • Coverage areas: 50% of US
          • Roaming: Depend upon carrier
          • Internet access: N/A
          • Type of data support: ???

      • Personal Communication Services (PCS)
        • CDMA and TDMA system.
        • GSM system.
          • Data Services: Dial-up, Circuit Switched
          • Coverage areas: 50% of US
          • Roaming: Depend upon carrier
          • Internet access: Yes
          • Type of data support: Dial-up at 9.6 Kbps

      • Satellite Systems
        • Iridium and Globalstar.

      • Wireless local Area Networks
        • IEEE 802.11 Wireless LAN Standard based systems, e.g., Lucent WaveLan.
        • Data Services: IP packets
        • Coverage areas: Offices, buildings, campuses
        • Roaming: Within deployed systems
        • Internet access: via LAN
        • Type of data support: Data at near LAN speed.

      • Packet Data Networks
        • ARDIS
        • RAM
        • Cellular Digital Packet Data (CDPD)

      • Private Networks
        • Public safety, UPS

    2. Wireless Data
      Data can be everywhere
      • Computers
      • LAN
      • Internet
      • Air!


  2. Wireless Data Management
    1. Characteristics and Assumptions
      • Limited and asymmetric bandwidth
      • Limited battery power on clients
      • Client machines have power managerment facility
      • A channel can be used either for on-demand, point-to-point service or broadcast service
      • Unlike Internet, bandwidth is a regulated and controlled system resource

    2. Performance Criteria
      • Access time:
        the time elapsed from the moment a MH issues a query to the moment the answer is received by the MH
      • Tune-in time:
        the time spent by a MH staying active in order to acquire the requested data
      • Goals:
        to make mobile client access time and tune-in time low while indexing efficiency high

    3. Wireless Data Dissemination
      • Many applications involve dissemination of data to large group of users, e.g., stock quotes, sport tickers and news delivery
      • Wireless dissemination of data is enabled by advances in Internet, data availability and wireless communications
      • An important issue: Scalability!

    4. Data push and pull
      Services in Mobile Environment

      • Broadcast Services
        • push-based
        • shared (serve an arbitrary number of users simultaneously)

      • On-Demand Services
        • pull-based
        • requests sent through an up-link channel
        • on-demand data delivered (access to any data item)
          • point-to-point communications (bi-directional)
          • through broadcast (by multiplexing with broadcast data)

    5. Broadcast Scheduling
      • Broadcast scheduling is to arrange the order and frequency of data items in a broadcast channel, in order to improve the client access time.
      • A common idea is to arrange the broadcast of data itmes based on their client access frequencies.

    6. Scheduling Approaches

      • Flat Broadcast
        • Broadcast every data item once per broadcast cycle

      • Multi-level broadcast
        • Data itmes are grouped together based on their client access frequencies
        • Groups of data itmes are broadcast based on their group weight
        • Data items in the same group are broadcast equally

    7. Channel Management

      • Exclusive on-demand channel allocation is desirable when the number of clients is small compared to the number of channels available

      • Exclusive broadcast channel allocation is great when a small number of data itmes are of interest to a large group of users

    8. Channel Allocation Medthods

      • With the limited and fixed bandwidth available in a cell, it's necessary to mix the broadcast and on-demand channels in order to compensate the weaknesses of the exclusive medthods and to improve the overall system performance

      • Mixed channel allocation methods
        1. Hybird Channel Allocation
        2. Dynamic Channel Allocation

    9. Caching
      • Main ideas: use cache to reduce access time
      • Caching policy: cache the most frequently accessed pages
      • Broadcast disk caching policy: pages are not arriving at the same time, so cache the important pages that are far away

    10. Indexing
      • To estimate the arrival of a requested data item, the client can select between broadcast and on-demand data services to reduce the access time
      • Cached schedule: the complete broadcast schedule is broadcast at the beginning of each cycle, each active client should monitor the channel for the schedule
      • Integrated signature


  3. Wireless Application Technologies

    1. Wireless Application Protocol (WAP)
    2. Wireless Application Environment (WAE)
    3. Wireless Markup Language (WML)


  4. Conclusion



回主頁  演講摘要  演講內容  相關網站