100 likes | 180 Views
Database Research – AG DBIS. Theo Härder haerder@informatik.uni-kl.de. bib. DOM tree example. b. o. o. k. title. a. u. t. h. o. r. a. u. t. h. o. r. author. p. r. i. c. e. year. id. l. a. s. t. f. i. r. s. t. l. a. s. t. f. i. r. s. t. l. a. s. t.
E N D
Database Research – AG DBIS Theo Härder haerder@informatik.uni-kl.de
bib DOM tree example b o o k title a u t h o r a u t h o r author p r i c e year id l a s t f i r s t l a s t f i r s t l a s t f i r s t T Data … 39,95 T T Dan Buneman T Sucio Peter Serge Abitebul 2 0 0 0 2 T T T T Meta-synchronization and XML Transactions • XUpdate – Concurrency control • node locks • navigation locks • predicate locks • Meta-lock requests • request shared node/level lock • request tree lock (shared, update, exclusive) • request edge lock (shared, update, exclusive) • XML Transaction Controller (XTC) as a testbed • comparison at the system level • unique research opportunity
XTCdriver DOM XTCconnection Browser FTP Client SAX L5 Interface Services Http Agent Ftp Agent DOM RMI SAX RMI API RMI XML Services XQuery Processor XML Manager XSLT Processor Node Processing Serv. Path Processing Transaction Services L4 Node Manager Lock Manager Access Services L3 Record Mgr Index Mgr Catalog Mgr Transaction Manager Propagation Control Deadlock Detector L2 Buffer Manager XTCserver File Services L1 Temp File Mgr I/O Manager Temporary Files Container Logs Transaction Log Container Files OS File System Native XML DBMSs (XTC) XTC – architectural overview- Reuse of the layer model is possible, but needs substantial adjustments and, in particular, new functionality in the higher layers - Focus on key mechanisms for efficient internal processing
SQL XML Databases – Query Processing Standard relational mapping Current solution (XTC) Vision ROX: relational over XML or on hybrid structures Language models DOM SAX XPath XQuery SQL declarative, set-oriented declarative, tree-structured, set-oriented generic navigational, node-oriented Access models Storage models relational hybrid native XML
XTC Project DFG
Technical term <command>bash</command> … <command> <filename> <programlisting> <programlisting> #!/bin/bash </programlisting> <filename>/bin/bash</filename> don’t <emph>bash</emph> him for that design decision Inline Markup in XML Retrieval Give me information about the shell “bash”! • Hope: semantic information about an element’s contents • But: element names are very specific • users: more abstract references • elements can be related in a graph structure Generalization based onElement Relationship Graph
Internet ISPj trans-parentcaching ISPl ISPk ISPi FW FW FW PTL LB RP/LB forwardproxycaching reverseproxycaching RP1 RPp P1 Pm … P11 P1k P: proxyRP: reverse proxy FW: firewallLB: load balancer BC1 BCn BCj BCt BCi SC2 SC1 C1 Cn Cj Ct Ci S1 S2 Client-to-Server Path Through the Internet • Web proxy servers • Response time (user-perceived delays), communication bandwidth, scalability of applications • Various types of caches, look-up of single Web objects • No support of the entire user-to-data path
CustomersWHERE Region=“west” DB Cache Application server DB Server Web server SQL Http Http DB Cache SQL Application server CustomersWHERE Region=“east” Browser requestsof clients Applicationlogic Front-End (FE)DB Server Back-End (BE)DB Server DFG Optimizing the User-to-Data Path • Constraint-based DB caching • Much more powerful than views • Processing of PSJ queries • Magic words: value complete, column complete, predicate complete • Idea can be applied to XML predicates and queries • New challenges • Evaluation of descriptive DB queries in the cache • Execution of update transactions?
Adaptive DBMSs A DBMS is (an important, but) only a component in a system • DBMSs have a very poor gain/pain ratio • Improve administration, extensibility, … • Are we on the right way? (XQuery, Web services, …) • Self-adaptation w.r.t. • Users and workloads • Resources, platforms, and environments • Information (representation, content, …) • Some progress for individual DBMS components • Rules of thumb (e. g. size of index pages) • Cache size adjustment to workloads? (KIWI?) • Online feedback control loop (observation, analysis, planning, reaction), e. g. load control
SQL, XQuery, Dyn. InfoFusion growing complexity observed needed simplification dependable systems engineering Dependable Adaptive Systems Much broader than self-tuning: dependable = self-* + trouble-free Dasmod • Feedback control loop? • Mostly non-functional properties • Intractable by mathematical models? • Dependable systems engineering • Highly componentized • Limited and relatively simple functionality/component • Component interfaces as narrow as possible • Design trends