We are going to present recently open sourced YoctoDB project — a small embedded Java-engine for extremely fast partitioned immutable-after-construction databases. We will briefly describe the architecture of indexing and search components, role and requirements on the search engine and our previous solution. Then we will dive into design and implementation of YoctoDB engine currently being used at Yandex.Auto and Auto.ru. In conclusion we will describe several Java pitfalls met along the way, the limitations of the approach and directions of future development.

The talk is targeted at software engineers and architects building high-load search systems and willing to optimize resource consumption and service latency. Design decisions and trade-offs are described to help implementing similar robust data processing pipelines.

The talk will be in English.

Vadim Tsesko

Lead of backend infrastructure team, Yandex.Classifieds

Vadim got into IT/CS in 2004. He has been developing high load distributed internet services at Yandex since 2011. Occasionally Vadim gives lectures and talks on Distributed storages and Parallel programming.

In his free time Vadim reads papers, solves Project Euler problems, plays PS4 with his wife or programs LEGO Mindstorms using Go.