Best practice when implementing nearby search with h3 indexing (write more or read more)

ScyllaDB
,
1

Hi guys, I’m trying to develop a system that requires to do some geospatial search (eg, find nearby users), so my first version of model of table has a set of h3 hashes of different resolutions of h3 cell ids stored (indexed, with TTL). When try to query nearby users, I make multiple queries for the disc of the cell to query (will generate 5-7 queries).

But I resemble reading somewhere that write queries in scylla db have less cost than read queries, so it came up to me that instead of storing only different resolutions of h3 cell of user’s current location, why not just store the whole cells in the disc and different resolutions (will increase the set size by 5-7x), but now only one query is required.

My question is does the latter version better than the first one or it really depends on the frequency of user’s location update and nearby search? Thanks!

2

You are correct that writes are cheaper in ScyllaDB than reads.
What you suggest makes sense to me, but it’s hard to know without more details. I’d try to create your data model and test it to see the actual performance and see if it’s suitable for your requirements.
You can use a tool like cassandra-stress to do the testing.
A good resource is the Basic Data Modeling lesson on ScyllaDB University.