Slide 1 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems Replication in Mobile Information Systems Hagen Höpfner Workshop Mobile Datenbanken und Informationssysteme Dortmund, October 2nd 2002
Slide 2 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems Agenda Introduction and motivation Basics Replication Process –Pre-processing –Copying the data –Updating the data and synchronisation Our approaches –Database server support for semantic replication –Mobile client support for semantic replication Conclusions and outlook
Slide 3 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems Introduction and Motivation Replication in distributed databases well known –Better performance –Better availability Replication in mobile information systems –Mobility and lightweight system –New demands What does Replication in Mobile IS mean?
Slide 4 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems
Slide 5 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems Basics Additional logical concepts devices –Database Server –Mobile (Database) Clients
Slide 6 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems The Replication Process
Slide 7 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems Pre-processing... On the Database Server: –Definition of the replica set Subset of the whole Database Content-based vs. Non-content-based –Allowing or Denying Updates... On a Base Station near the MH –Caching Only new Data from DS –Profiling... On the Mobile Client –Caching Reuse of already replicated Data Classical Approaches vs. Semantic Caching... On the Database Server: –Definition of the replica set Subset of the whole Database Content-based vs. Non-content-based –Allowing or Denying Updates... On a Base Station near the MH –Caching Only new Data from DS –Profiling... On the Mobile Client –Caching Reuse of already replicated Data Classical Approaches vs. Semantic Caching... On the Database Server: –Definition of the replica set Subset of the whole Database Content-based vs. Non-content-based –Allowing or Denying Updates... On a Base Station near the MH –Caching Only new Data from DS –Profiling... On the Mobile Client –Caching Reuse of already replicated Data Classical Approaches vs. Semantic Caching
Slide 8 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems Copying the Data Disk-File-Based –Requires additional hardware Not usable in most cases Agent-Based Network-Protocol-Based –Can also support files Broadcast –Requires separate uplink channel for querying
Slide 9 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems Updating and Synchronisation Data centric –Holding different versions of data –E.g. Multiversion Reconciliation Transaction centric –Tentative transactions vs. Base updates –E.g. Two-Tier Replication –E.g. Bayou: Per-write dependency checks Per-write merge procedures
Slide 10 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems Our Pre-processing Approaches Using characteristics of mobile information systems for minimizing wireless data transmissions
Slide 11 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems Semantic Replication: DS Fragmentation of the database on DS –Fragment=Set of Data + Server Extensions Queries may contain additional context- descriptions Only related fragments become replicated Example
Slide 12 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems Semantic Replication: MC (1) Usage of typical query structure for semantic caching on mobile clients –Content-based queries –A lot of small queries with small but exact answers Usage of a Trie for indexing cached data –Edges represent predicates
Slide 13 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems Semantic Replication: MC (2)
Slide 14 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems Conclusions Replication in mobile information systems differs from replication in distributed databases in: –Intensions –Requirements –Techniques Excurs: Our actual works on mobile information systems
Slide 15 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems Outlook Support for local joins Detailed specification of semantic comparison of queries Representation of constraints etc. (The View Update Problem) Integration of update mechanisms (The Offline Update Problem)
Slide 16 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems Discussion
Slide 17 School of Computer Science Hagen Höpfner Replication in Mobile Information Systems Institute of Technical an Business Information Systems Example: Server Site AMO (date, time, event, event_type) CINEMAXX (id, room#, fsk, date, time, movie, genre) FACTORY (date, time, admittance, band, charge, genre) Fragmentation function: –F c md =(f c md, v l md ) –f c md =select Date, Admittance as Time, Band, Charge, Genre from FACTORY union select Date, Time, Event, NULL, NULL from AMO where event_type like '%concert%'; back