179 questions
No questions match those filters.
← All questions
Retrieval & EmbeddingsHard· DiskANN· Latency· Performance Tuning
Your DiskANN p99 is 12 ms against a 5 ms budget. What d...
This is one of the questions in the full AI/ML interview bank. Pro unlocks all 1789 questions; Premium includes the same bank plus the highest daily Practice limit.
See plansDiskANN latency is driven by the number of I/O rounds. Increasing beam width reduces the number of rounds but increases the number of reads per query, which lowers the QPS ceiling. It is critical to check if vectors cross sector boundaries, as this causes silent performance degradation. Only after these optimizations should you consider rebuilding the graph with a higher alpha parameter.