Description
Accomplishments * Designed, implemented and optimized system for monitoring of universe of company database servers to troubleshoot performance problems, proactively monitor environment performance, and develop sophisticated statistical analysis of historical data and quantitative charge-back models. System processes ~200,000,000 records per day. Project was presented on Sybase Techcast 2003 and ISUG July 2004 session. * Fostered methodic for optimization of sophisticated SQLs which combine joins, unions and aggregations. Methodic is proved on multiple patterns and provide strong run time reductions. (For BEAR STEARNS and JPMORGAN). * Developed generic approach for loading of huge amount of records into database in a loop of moderate batches. Project allows combining optimal set based performance with best quality of load - all good records loaded even if batch includes some bad records causing SQL to fail - and may be applied to any SYBASE or ORACLE load processing. (For BEAR STEARNS and JPMORGAN). * Created original numerical methods for database financial applications, which include tabulating of special non-linear functions. Method produces set of data with single SQL call. (For GOLDMAN SACHS and BEAR STEARNS). * Investigated Bloom filters usage in Exadata and developed practical suggestions improving its efficiency (For JPMORGAN). * Investigate customized alternatives to vendor's functionality (analytical functions, hierarchical queries, super-aggregates) and prove benefits of former in many cases. * Developed generic script which looked for "bad indexes" - indexes that don't improve select performance. Script is based on reading and analyzing system tables only and is applicable to both Oracle and Sybase (For BEAR STEARNS and JPMORGAN).
Accomplishments
Highlights:
Keywords
