No Mean Feat: Upgrading a Customized Solr to Upstream Solr
06-19, 14:50–15:10 (Europe/Berlin), Maschinenhaus

Learn how the News Search Infrastructure Team at Bloomberg migrated from a customized implementation of Apache Solr to the upstream Apache Solr


Technology upgrades are a key pillar of software infrastructure. However, upgrading search and information retrieval systems is a complex task. At Bloomberg, we had extended the open source Apache Solr implementation with in-house patches to customize it for our use cases. This made upgrading to a newer version quite challenging. But, when you have close to a billion documents that are used by major financial institutions across the world, you cannot afford any mistakes.

Learn how the News Search Infrastructure Team at Bloomberg migrated from a highly customized implementation of Apache Solr to the upstream Apache Solr, while also making sure that the quality, correctness, and performance of the system was not affected. You will learn about the different strategies we used before, during, and after the migration to make the upgrade transparent to our internal users, all while serving millions of requests everyday.

Shikhar is a Software Engineer on the News Search Infrastructure Engineering team at Bloomberg in London. He has worn multiple hats in his professional career, from developing ETA prediction machine learning models for startups in India to developing low latency, financial market data systems at Bloomberg. He recently started dabbling with Apache Solr and has fallen in love with it