- Owned the dev side of the back end of SQL Server Database Tuning Advisor . DTA analyzes a representative workload consisting of queries and updates that run against the database engine and recommend appropriate changes to physical design for the workload. Recent work include tuning workload from the engine query plan cache, recommending compressed indexes, filtered indexes and extending existing index selection algorithms to exploit multiple processors.
- Implemented mechanisms for a multi-tenant, shared buffer pool governance in SQL Engine. This involved adapting novel algorithms to modify storage engine’s traditional page replacement policy (LRU-K based) to different SLA aware replacement policies.
- I have worked on workload driven index defragmentation in the database engine. We developed low overhead techniques to identify and defragment a logical range rather than the whole physical B-Tree based on the workload, making the index reorganize/rebuild operation amenable to online and zero admin settings.
- In the past I have worked on developing a framework for dynamic profiling and static analysis of database centric applications. We adapted profiling techniques and traditional program analysis (using extensible compiler framework Microsoft Phoenix) by leveraging understanding of call level data access APIs.
I work as a Software Design Engineer in the DMX group.
As a member of the SQLVM team, I designed and implemented the SQL Azure I/O resource governor mechanism in the database kernel. Some of the interesting challenges we had to deal with included a new I/O scheduling mechanism aligned with the SQL OS scheduler, scaling to thousands of database tenants on one node and accommodating more than one, possibly heterogeneous throughput capacity storage media.