Mapping Data to Queries Martin Hentschel Systems Group, ETH Zurich

Martin Hentschel/Systems Group, ETH  “…, but the real advantage of XML is precisely that it allows you to go from Point A to destinations unknown.” -- Larry O’Brien, Microsoft 2

Martin Hentschel/Systems Group, ETH 3 Goals  Integrate data from various data feeds  Light-weight  Easy to use  Fast

Martin Hentschel/Systems Group, ETH 4 Goals  Integrate data from various data feeds  Light-weight Mapping rules  Easy to use Based on common language (XQuery)  Fast Implements research ideas (YFilter)

Martin Hentschel/Systems Group, ETH Targets  Health care  Electronic health records (Health Level 7)  Finance  Exchange of financial data (xBRL)  Web services  News feeds  Weather Every domain which uses several data sources 5

Martin Hentschel/Systems Group, ETH Example  Find the most powerful car 6 Ford 130 Ford 130 VW Golf 150 VW Golf 150

Martin Hentschel/Systems Group, ETH Example  Find the most powerful car 7 Ford 130 Ford 130 VW Golf 150 VW Golf 150 datenis-adb; autois-acar; psis-ahp; datenis-adb; autois-acar; psis-ahp;

Martin Hentschel/Systems Group, ETH Example  Find the most powerful car  Apply standard XQuery 8 Ford 130 Ford 130 VW Golf 150 VW Golf 150 datenis-adb; autois-acar; psis-ahp; datenis-adb; autois-acar; psis-ahp; let $max := max(//hp) for $car in //car where $car/hp = $max return $car let $max := max(//hp) for $car in //car where $car/hp = $max return $car

Martin Hentschel/Systems Group, ETH Example  Find the most powerful car  Apply standard XQuery 9 Ford 130 Ford 130 VW Golf 150 VW Golf 150 datenis-adb; autois-acar; psis-ahp; datenis-adb; autois-acar; psis-ahp; let $max := max(//hp) for $car in //car where $car/hp = $max return $car let $max := max(//hp) for $car in //car where $car/hp = $max return $car VW Golf 150 VW Golf 150  Result

Martin Hentschel/Systems Group, ETH Usage Scenarios  Continuous query processing 10 DSMS Querie s Rules Streaming Input Events Streaming Output Events

Martin Hentschel/Systems Group, ETH Usage Scenarios  Publish/subscribe systems 11 Rules PublishersSubscribers Enhanced Broker Enhanced Broker Data Subscriptions Data

Martin Hentschel/Systems Group, ETH Usage Scenarios  Data integration 12 Rules Source 1 Company‘s Data Store Data Source 2 Source x Homogeneous Data Handler Data Handler

Martin Hentschel/Systems Group, ETH The Is-A Rule  Map XML elements  Expresses a substitutability relationship  Like in object oriented design  Use the car wherever vehicles are expected  It follows //vehicle also returns car elements  Returned as car  Not transformed into vehicle  Consistent with OO-approach 13 car is-a vehicle;

Martin Hentschel/Systems Group, ETH The Is-A Rule  Map path expressions  XPath path expressions  Left hand side may include predicates 14 german/car is-a auto; auto is-a german/car; german/car is-a auto; auto is-a german/car; < 100] is-a slow/vehicle; < 100] is-a slow/vehicle;

Martin Hentschel/Systems Group, ETH The Is-A Rule  Specify contexts  Element names could be used differently in different contexts  Scope applicability of rules  Further refinement 15 car in is-a auto; car in is-a auto;

Martin Hentschel/Systems Group, ETH The Is-A Rule  Element construction  Map elements  Transform data, e.g. for  Integration of very diverse data 16 auto as $a is-a {$a/ps * 0.74} ; auto as $a is-a {$a/ps * 0.74} ; Ford 100 Ford 100 VW Golf 150 VW Golf 150

Martin Hentschel/Systems Group, ETH Implementation  Several possibilities  MDQ approach -Native approach, novel MDQ data model -Allows lazy execution  Query rewrite -E.g. //(car | auto | vehicle |...) -Does not scale  Data translation -Translate input data -Big overhead 17

Martin Hentschel/Systems Group, ETH MDQ Data Model  Classical XML tree model 18 Golf 150 Golf 150 auto psname „Golf“„150“ daten

Martin Hentschel/Systems Group, ETH MDQ Data Model  MDQ data model  Move names from nodes to edges 19 Golf 150 Golf 150 auto psname „Golf“„150“ daten

Martin Hentschel/Systems Group, ETH MDQ Data Model  Application of mapping rules 20 Golf 150 Golf 150 auto ps name „Golf“„150“ daten daten is-a db; auto is-a car; ps is-a hp; daten is-a db; auto is-a car; ps is-a hp; db car hp

Martin Hentschel/Systems Group, ETH Lazy Evaluation, YFilter  Built from left hand side of rules  Non-deterministic finite state machine  Main idea:  Evaluate XQuery program  Iterate through data model  Report to YFilter  Apply rules only when reaching an accepting state 21 R1: daten is-a db; R2: auto is-a car; R2: ps is-a hp; R1: daten is-a db; R2: auto is-a car; R2: ps is-a hp;  * daten auto ps R1 R2 R3

Martin Hentschel/Systems Group, ETH Experiment: Throughput  Complex query (multiple scans, joins)  QR: too many unions, DT: overhead of translation 22

Martin Hentschel/Systems Group, ETH Experiment: Throughput  Simple query  Less unions for QR, DT: still overhead of translation 23

Martin Hentschel/Systems Group, ETH Experiment: Throughput  1 input message, bundle of queries evaluated at once  QR: even more unions, DT: less overhead, only transforms input message once 24

Martin Hentschel/Systems Group, ETH Again: Advantages  Performance  Novel data model, lazy execution  Light-weight  Mappings rules are small units  Extensibility  Add more rules as new sources are adopted  Flexibility  Complex mappings through element constructors 25

Martin Hentschel/Systems Group, ETH The End  Visit our website, LIVE DEMO!   Write us, please!  26