When I read this question, I have to wonder if you're approaching this as a general solution to replace search, or if you're intending on using SAI as a means to filter within a partition. It's designed to solve the latter.
The overhead of an SAI query scales with the number of SSTables you read. If you do a global query, you're hitting every SSTable in your cluster. That means a 6 node cluster has twice the overhead as a 3 node, and it gets worse linearly as you scale up. It's really not meant to solve this type of problem, and you will encounter serious issues at scale that can't be fixed easily, or inexpensively. Jon On Mon, May 12, 2025 at 7:38 AM Manish Sharma <mannu2...@gmail.com> wrote: > Hi, > We are planning to remove Lucene due to most of the queries we are using > now supported via SAI, however we have few queries which are requiring > Facets and looking for information on how to address them. Also, if there > is any roadmap to include it. > > Regards > Manish Sharma >
