Kiva Teknoloji, a Turkish company developing cloud business applications since 2009, manages a SaaS platform where each client maintains a separate database. Due to the platform's low-code/no-code nature, database structures change frequently as clients customize fields and business processes.
To address search performance, Kiva integrated Manticore Search as a dedicated layer alongside MySQL. Previously, the company relied on MySQL's built-in full-text search, which consumed significant CPU and memory resources, impacting overall application performance.
The current architecture uses a distributed Manticore table for each client, combining periodically rebuilt tables with real-time (RT) tables for live updates. Manticore returns record IDs, while the application fetches full data from MySQL, preventing data duplication.
Kiva does not deploy dedicated servers for Manticore, running the engine on existing application or database replica servers instead. This approach has minimized infrastructure costs while freeing up resources on the primary MySQL databases.
Looking ahead, Kiva is exploring the implementation of hybrid search capabilities, combining full-text and vector search within the existing Manticore infrastructure to support future AI-driven application features.
