Speed-of-light faceted search via Oracle In-Memory option

Edward 1-4

The talk is real-life story about search and findings for high-performance faceted search engine of one of our customers.

Client's application struggled from performance degradation for either data ingestion and ad-hoc search Before In-Memory approach and it is in a perfect shape now.

The talk will briefly describe all former faceted search architectures and options tested before in-memory and why Oracle In-Memory was choosen.

I show how various In-Memory Oracle features were used, which of them are whether good to use or not. In-memory compression, partial in-memory allocation, columnar storage will be investigated in detail.

Pieces of real queries and performance comparison from production system will be shown.

Possible maintainance and DBA risks will be highlighted as well.

The talk is intended for tech architects, DBA who would like to have a look how In-Memory solutions should be tested and put in production in terms of limited time resource.


Profile picture for user shtock
Performance architect
Alexander joined DataArt in 2015. Prior to his work at DataArt, he was engaged in the development of large data warehouses (30+ Tb) in various subject areas. Alexander is a certified Oracle specialist with 12 years of experience working with DBMS and is an expert in MPP databases. As of now, he is still engaged in the development of data lakes as a whole and in particular searching for new approaches to this process.