Database Change Notifications Primitives for Efficient Database Query

Database Change Notifications: Primitives for Efficient Database Query Result Caching Cesar Galindo-Legaria, Torsten Grabs, Christian Kleinerman, Florian Waas SQL Server Engine Microsoft Corp. Redmond, WA USA VLDB 2005, Trondheim, Norway http: //www. microsoft. com/sql/2005/

Motivation Clients HTTP Mid-tier cache: Web server farm SELECT price FROM books WHERE title = ‘T-SQL’ DBS Title Price T-SQL Title 10. 00 Price T-SQL 20. 00 10. 00 HTTP 15. 00 UPDATE books SET price = 20. 00 WHERE title = ‘T-SQL’ • Three-tier architecture for web applications • Time-to-life based approaches for caching of read-mostly dynamic results at mid-tier for performance reasons • Problem: Outdated results in case of database updates 2

Demo Highlights Clients HTTP Mid-tier cache: Web server Query Plan farm SELECT Update Plan DBS Title Price T-SQL 20. 00 10. 00 HTTP 15. 00 Title Price T-SQL 10. 00 20. 00 Change Notification: - cache re-run query, or - send delta UPDATE • “Non-materialized” indexed view for change detection – Query plan rewrite to register cached query results – Update rewrite to notify caches of changes in case of nonempty materialized view maintenance delta stream – Note: No DBMS installation necessary at the mid tier 3

? Questions ? Come and see the Demo!!! 4