Imielinski, Tomasz, B.R. Badrinath. "Mobile Wireless Computing: Challenges in Data Management" 1. INTRODUCTION Scenario: A scenario in the near future could include millions of users carrying portable computers ranging from PDAs to laptops, all with wireless connections to information networks and running mail enabled applications and information services. News and information will be delivered and filtered according to individual user profiles. Being able to stay connected will stimulate more collaborative computing. Local yellow pages could deliver local information such as movies playing at local theaters or merchandise on sale at the local supermarket. Architecture: Mobile support stations (MSS) are fixed hosts which have a wireless interface to support mobile hosts. MSSs will provide commonly used software so that mobile hosts can downloaded them or run them remotely on the MSS. Each mobile user will be associated with a home MSS, which stores user profile, login file, access rights, private files. Mobile hosts could be dumb terminals (with no disk) and limited RAM or walkstations (which will have disks and more powerful processors). Xerox Tabs are wallet sized devices with a display and can communicate using infrared. Mobile computers have limited battery life, will be frequently disconnected, may wake up in a distant location far away from home, and may cross the border between two different cells while active (handoff). Major challenges in data management: 1. Mobility Management and Scalability 2. Bandwidth Management 3. Energy Management 2 Gen. Architectures 2.1 Networks Cellular architecture wide area moves poor scalability no data transmission low bandwidth Wireless LANs limited range no wide area moves Wide Area Wireless Networks no ubiquitous network connectivity poor scalability Paging networks receive only very low bandwidth Satellite networks have been proposed for paging, fax, and messaging, e.g. Qualcomm's Globalstar and TRW's Odyssey Bit rates Current Future Wireless LAN 1 to 2 Mbps 10Mbps Infrared 19.2kbps to 1Mpbs Wireless WAN 19.2kbps (SMR and CDPD) Tariffs $30 to 100 flat fee Data Services $0.35/50 char message $1/100 char message Cellular $0.15 to $0.60/minute 2.2 Palmtops low speed CPUs, limited memory, short battery life 3 Information Services for Mobile Users Ubiquitous network scenario. 4 Mobility and Scale mobility is an important new component that will have far reaching consequences for systems design. 4.1 Location management: mobility of clients Home address is stored on a distributed directory server. Finding out the current location is still an open problem. Fundamental tradeoff: searching or informing? Current cellular location management and proposed Internet mobility: when a user moves, he informs the home server of his new address. All packets sent to home server are redirected to mobile address. Disadv. can't handle global moves. Improvement: Query local servers before contacting home server. 4.2 Distributed Systems (Configuration management: mobility of resources) Should resources be static or mobile? 4.3 Scale Total volume of transactions is predicted to be 1 order of mag. higher than current network capacities. Scale may affect location management, resource (configuration) managment, cell diameter, (more users means smaller and smaller cells and more frequent handoffs), and heterogeneity. 5. Bandwidth and Energy Management Data broadcasting Interactive/On-demand. To maximize query capacity, broadcast most-frequently accessed itme.s To save energy, broadcast a channel index so that the mobile unit can wake up only when necessary. Information should be organized so that Access Time and Tuning Time are minimized. Index information will increase overall broadcast size, but will decrease tuning time and thus energy consumption. 6. What is affected 6.1 Large impact Distributed Data Management connections are no longer fixed dynamic replication of data and services Query Processing location information may be approximate broadcasting is suitable for a large number of clients Security and Integrity where should profiles be stored? Black pages Interface design Data entry limitations Data output limitations 6.2 Some impact Transaction processing Form based transactions let users fill in fields of a form Long disconnections locking and commit protocols need to be redefined Integrity and Recovery how to maintain referential integrity without having to contact mobile hosts all the time? elective failures introduce new problems for recovery 6.3 Little or no impact Data Modeling require new techniques for user's profile specification must describe mobility pattern for the user Little impact on data manipulation language design, except for special constructs for spatial queries, active rules and triggers for spatial data.