Going beyond the realm of text search, SaaS applications commonly offer users the ability to search and filter across many different fields in the data. This type of search feature can be challenging to build, especially when results need to reflect real-time data and user queries are not known in advance. It may require queries to be rewritten in a way the database supports or data to be restructured or indexed in specific ways to ensure good query performance.

In this talk, we explore different ways of building real-time search functionality into SaaS applications, using databases like Elasticsearch and Rockset. We’ll draw upon our experience advising development teams on this process and discuss the importance of being able to query in different ways without having to reshape your data. Topics to be discussed include:

• Indexing strategies- Indexing is key to delivering the low query latency needed by real-time search. Inverted indexes are generally used for text search and are very helpful for implementing search functionality in applications. Beyond just inverted indexes, another approach is to index data in multiple ways, also known as converged indexing, so different types of queries can benefit from indexing.
• SQL or domain-specific language- SQL allows for greater familiarity among developers and support for a full range of aggregation queries. In comparison, a domain-specific language may present users with more limitations in the functionality and ecosystem it supports.
• Support for joins- Developers may need to work around a lack of join support by denormalizing the data at ingest or writing an application-side join. The ability to do joins allows developers greater flexibility when analyzing their data.