Performance and Optimisation Interview Questions | Qdrant | NextRound

1Your Qdrant search endpoint's p50 latency looks fine, but p99 latency is very high. What are the most likely causes to investigate first?Advanced

2Why does Qdrant recommend disabling indexing (or raising the indexing threshold) during a large bulk import, then re-enabling it afterward?Intermediate

3What is the purpose of the indexing_threshold setting, and how does it affect small versus large collections differently?Advanced

4How does GPU-accelerated indexing change the economics of re-indexing a large, frequently-updated collection?Advanced

5What is incremental HNSW indexing, and why does it matter for upsert-heavy workloads?Advanced

6How would you reduce query latency for a collection that must remain on-disk due to its size, without moving the whole collection into RAM?Advanced

7What is the effect of increasing the number of search threads/parallelism on a single node with limited CPU cores?Intermediate

8How would you benchmark whether a proposed quantization configuration is worth the accuracy trade-off for your workload?Advanced

9What's the difference between scaling Qdrant vertically (bigger node) and horizontally (more shards/nodes), and when does horizontal scaling stop paying off?Advanced

10Two teams store the same 50-million-vector collection - one keeps it fully in memory, one on disk with quantization. What operational differences should each expect?Advanced

11Why can moving the payload storage engine on-disk versus in-memory have a bigger impact on filtered-search latency than the vector storage location?Advanced

12How would you decide, for a specific collection, whether to enable quantization with rescoring versus simply moving vectors on-disk without quantization?Advanced

13What memory overhead does the HNSW graph itself add on top of the raw vector data, and why does that matter when planning RAM for an in-memory collection?Advanced